Glandular architecture is currently the basis for the Gleason grading of prostate biopsies. To visualize and computationally analyze glandular features in large 3D pathology datasets, we developed an annotation-free segmentation method for 3D prostate glands that relies upon synthetic 3D immunofluorescence (IF) enabled by generative adversarial networks. By using a fluorescent analog of H and E (cheap and fast stain) as an input, our strategy allows for accurate glandular segmentation that does not rely upon subjective and tedious human annotations or slow and expensive 3D immunolabeling. We aim to demonstrate that this 3D segmentation will enable improved prostate cancer prognostication.