Distinguishing Renal Cell Carcinoma From Normal Kidney Tissue Using Mass Spectrometry Imaging Combined With Machine Learning

Abstract

Purpose: Accurately distinguishing renal cell carcinoma (RCC) from normal kidney tissue is critical for identifying positive surgical margins (PSMs) during partial and radical nephrectomy, which remains the primary intervention for localized RCC. Techniques that detect PSM with higher accuracy and faster turnaround time than intraoperative frozen section (IFS) analysis can help decrease reoperation rates, relieve patient anxiety and costs, and potentially improve patient outcomes.

Materials and methods: Here, we extended our combined desorption electrospray ionization mass spectrometry imaging (DESI-MSI) and machine learning methodology to identify metabolite and lipid species from tissue surfaces that can distinguish normal tissues from clear cell RCC (ccRCC), papillary RCC (pRCC), and chromophobe RCC (chRCC) tissues.

Results: From 24 normal and 40 renal cancer (23 ccRCC, 13 pRCC, and 4 chRCC) tissues, we developed a multinomial lasso classifier that selects 281 total analytes from over 27,000 detected molecular species that distinguishes all histological subtypes of RCC from normal kidney tissues with 84.5% accuracy. On the basis of independent test data reflecting distinct patient populations, the classifier achieves 85.4% and 91.2% accuracy on a Stanford test set (20 normal and 28 RCC) and a Baylor-UT Austin test set (16 normal and 41 RCC), respectively. The majority of the model's selected features show consistent trends across data sets affirming its stable performance, where the suppression of arachidonic acid metabolism is identified as a shared molecular feature of ccRCC and pRCC.

Conclusion: Together, these results indicate that signatures derived from DESI-MSI combined with machine learning may be used to rapidly determine surgical margin status with accuracies that meet or exceed those reported for IFS.

FIG 1.

DESI-MS analysis of normal kidney and RCC tissues, specifically pRCC, shows that normal tissues have higher abundance of m/z 317.2249 (12S-HEPE) compared with RCC tissues. (A) Representative DESI mass spectra of normal kidney tissue and the respective DESI-MS ion images of metabolites m/z 317.2249 (12S-HEPE) and m/z 281.2484 (oleic acid). (B) Representative DESI mass spectra of pRCC tissues and the respective DESI-MS ion images of 12S-HEPE and oleic acid. DESI-MS, desorption electrospray ionization mass spectrometry; FA, fatty acid; NL, normalization; PG, glycerophosphoglycerol; pRCC, papillary RCC; RCC, renal cell carcinoma; 12S-HEPE, 12S-hydroxy-5Z,8Z,10E,14Z,17Z-eicosapentaenoic acid.

FIG 2.

(A) AUC-ROC analysis compares the performance of the multinomial lasso model across training and test sets and finds stable performance. (B) Comparison of lasso selected m/z peaks from Stanford training to both Stanford and Baylor-UT Austin test data sets indicates that the majority of peaks either associated with normal or cancer tissues maintain consistent trends. (C) Comparison of the arachidonic acid (m/z 303.23) relative intensity across normal, ccRCC, pRCC, and chRCC finds this metabolite is elevated in normal tissues compared with ccRCC and pRCC but lower in normal tissues compared with chRCC levels. AUC-ROC, area under the curve-receiver operating characteristic curve; ccRCC, clear cell RCC; chRCC, chromophobe RCC; CV, cross-validation; pRCC, papillary RCC; RCC, renal cell carcinoma.

FIG 3.

(A) Dendogram from hierarchical clustering of Stanford test (S) and Baylor-UT Austin test tissues (B-UT) is shown. Dot colors correspond to RCC subtype, and perforated boxes are drawn around leaves where tissues of distinct RCC histological subtypes cluster together. (B) Comparison across tissues of what proportion of Stanford test tissues is most similar to a Baylor-UT Austin test tissue of the same subtype. The number of tissues is shown in parentheses, where 28 total tissues from the Stanford test set were evaluated (18 clear cell, eight papillary, and two chromophobe). RCC, renal cell carcinoma.

FIG 4.

DESI-MS analysis shows representative spectra from normal, pRCC, ccRCC, and chRCC tissues. Spatial maps corresponding to arachidonic acid (m/z 303.23) and 11-dehydro-TXB2-d4 (m/z 371.28) suggest elevation of these metabolites in normal compared with RCC tissues. Histology of RCC subtypes was confirmed by H&E staining tissue section from the same cases (shown on right at 40× magnification). ccRCC, clear cell RCC; chRCC, chromophobe RCC; DESI-MS, desorption electrospray ionization mass spectrometry; H&E, hematoxylin and eosin; NL, normalization; pRCC, papillary RCC; RCC, renal cell carcinoma.