Spatial metabolomics informs the use of clinical imaging for improved detection of cribriform prostate cancer

Abstract

Cribriform prostate cancer (crPCa) is associated with poor clinical outcomes, yet its accurate detection remains challenging due to the poor sensitivity of standard-of-care diagnostic tools. Here, we use untargeted spatial metabolomics to identify fatty acid biosynthesis as a key metabolic pathway enriched in crPCa epithelium. We also show that imaging tumor lipid metabolism using [1-¹¹C]acetate PET/CT and proton magnetic resonance spectroscopy differentiates cribriform from noncribriform intermediate-risk prostate cancers in two prospective patient cohorts. These findings support the feasibility of using clinical metabolic imaging techniques as adjunctive tools for improving crPCa detection in clinical practice, with prospective studies in larger cohorts warranted to obtain definitive results.

Keywords: MRI; cancer metabolism; nuclear medicine; prostate cancer; spectroscopy.

Fig. 1.

Clinical metabolic imaging for crPCa detection in surgical patients. (A) Study flowchart including the constituent subcohorts with matching clinico-pathologic characteristics. (B) Comparison of ¹H-MRI tumor-to-urine ADCratio and digital pathology-derived epithelial cell density between cribriform and noncribriform GS7 tumors in the surgical cohort. (C) Metabolic pathway enrichment analysis of cribriform (red) and noncribriform (yellow) epithelial regions from the spatial metabolomics subcohort, with top three pathways labeled for each morphological entity; for the same plot featuring individual metabolites, see SI Appendix. (D) Representative hematoxylin and eosin (H&E) and DESI-MSI slides from specimens containing cribriform and noncribriform morphologies derived from the same patient from the spatial metabolomics subcohort; scale bars denote 2 mm. (E) Comparison of the percentage of cells with strongly positive MCT1 and FASN immunoexpression between cribriform and noncribriform tumors from the IHC subcohort. (F) Representative H&E and IHC images from the IHC subcohort; scale bars denote 50 μm. (G) Comparison of [1-¹¹C]acetate SUV_bw between cribriform and noncribriform tumors from the [1-¹¹C]acetate PET/CT subcohort; yellow dots denote SUV_bw values for the two cribriform lesions misclassified at diagnostic biopsy. (H) Representative whole-mount H&E, ¹H-MRI, and [1-¹¹C]acetate PET/CT images from a patient harboring [1-¹¹C]acetate PET/CT-visible cribriform lesion in the left peripheral zone (white arrow) and [1-¹¹C]acetate PET/CT-occult noncribriform transition zone tumor.

Fig. 2.

Clinical fat-water ¹H-MRSI for crPCa detection in patients undergoing MRI-targeted biopsy. (A) Cohort description. (B) Comparison of ¹H-MRI tumor-to-urine ADCratio and ¹H-MRSI fat fraction between areas of histologically benign prostate, along with GS6, GS7 noncribriform, and GS7 cribriform tumors. (C) Representative T₂-weighted images (T2WI) from patients with noncribriform (Top) and cribriform (Bottom) GS7 tumors with black, yellow, and red dotted areas denoting benign, noncribriform, and cribriform regions-of-interests (ROIs) used for extracting ¹H-MRI tumor-to-urine ADCratio and ¹H-MRSI fat fraction data. The panel also presents fused T2WI and ¹H-MRSI fat fraction maps from the same patients, along with representative 2 × 2 spectra from voxels within the outlined benign and tumor ROIs; lipid peaks are denoted by black arrows, with boxes in the right top corner of spectra presenting their magnified images. To aid visualization, these spectra have thicker lines and are black in color compared to the raw spectra presented in SI Appendix.