Gleason grading of histological images is important in risk assessment and treatment planning for prostate cancer patients. Much research has been done in classifying small homogeneous cancer regions within histological images. However, semi-supervised methods published to date depend on pre-selected regions and cannot be easily extended to an image of heterogeneous tissue composition. In this paper, we propose a multi-scale U-Net model to classify images at the pixel-level using 224 histological image tiles from radical prostatectomies of 20 patients. Our model was evaluated by a patient-based 10-fold cross validation, and achieved a mean Jaccard index of 65.8% across 4 classes (stroma, Gleason 3, Gleason 4 and benign glands), and 75.5% for 3 classes (stroma, benign glands, prostate cancer), outperforming other methods.

Men with high-grade prostate cancer (HGPC) are at greatest risk of disease progression. Clinical risk factors associated with castration-resistant prostate cancer (CRPC), metastases, and prostate cancer-specific mortality (PCSM) were identified in a contemporary HGPC cohort. Clinical data was collected from men diagnosed with Gleason sum (GS) ≥8 at the Greater Los Angeles Veterans Affairs (GLA-VA) Healthcare System between 2000 and 2013. Multivariable competing risks regression analyses assessed progression to CRPC, metastases, and PCSM within three treatment strata. The cumulative incidence of disease progression was calculated at 2, 5, and 10-year time points. Review of 2149 prostate cancer cases yielded 322 with HGPC. Median survival times for cancer-specific and overall mortality were significantly shorter in men treated with primary androgen deprivation therapy (ADT) (P = 0.0002 and P < 0.0001). Multivariable analyses revealed that clinical stage N1, GS 10, and treatment with primary ADT were significantly associated with increased risk of CRPC, metastases, and PCSM. Significant differences in these outcomes were not observed in men treated with radical prostatectomy (RP) when compared to those treated with radiation therapy combined with short-term ADT (XRT-ADT). Ten-year event rates of progression to CRPC, metastases, and PCSM, for men treated with primary ADT, were 45.5%, 25.4%, and 25.1%, respectively. In conclusion, GS 10 and lymph node involvement, as well as primary ADT treatment in men with HGPC was associated with increased risk of CRPC, metastases, and PCSM. Curative-intent treatment with RP or XRT-ADT is associated with reduced progression rates and death in men with HGPC.

Prostate cancer is the most common and second most deadly form of cancer in men in the United States. The classification of prostate cancers based on Gleason grading using histological images is important in risk assessment and treatment planning for patients. Here, we demonstrate a new region-based convolutional neural network framework for multi-task prediction using an epithelial network head and a grading network head. Compared with a single-task model, our multi-task model can provide complementary contextual information, which contributes to better performance. Our model is achieved a state-of-the-art performance in epithelial cells detection and Gleason grading tasks simultaneously. Using fivefold cross-validation, our model is achieved an epithelial cells detection accuracy of 99.07% with an average area under the curve of 0.998. As for Gleason grading, our model is obtained a mean intersection over union of 79.56% and an overall pixel accuracy of 89.40%.

Immunohistochemistry (IHC) is used to guide treatment decisions in multiple cancer types. For treatment with checkpoint inhibitors, programmed death ligand 1 (PD-L1) IHC is used as a companion diagnostic. However, the scoring of PD-L1 is complicated by its expression in cancer and immune cells. Separation of cancer and noncancer regions is needed to calculate tumor proportion scores (TPS) of PD-L1, which is based on the percentage of PD-L1-positive cancer cells. Evaluation of PD-L1 expression requires highly experienced pathologists and is often challenging and time-consuming. Here, we used a multi-institutional cohort of 77 lung cancer cases stained centrally with the PD-L1 22C3 clone. We developed a 4-step pipeline for measuring TPS that includes the coregistration of hematoxylin and eosin, PD-L1, and negative control (NC) digital slides for exclusion of necrosis, segmentation of cancer regions, and quantification of PD-L1+ cells. As cancer segmentation is a challenging step for TPS generation, we trained DeepLab V3 in the Visiopharm software package to outline cancer regions in PD-L1 and NC images and evaluated the model performance by mean intersection over union (mIoU) against manual outlines. Only 14 cases were required to accomplish a mIoU of 0.82 for cancer segmentation in hematoxylin-stained NC cases. For PD-L1-stained slides, a model trained on PD-L1 tiles augmented by registered NC tiles achieved a mIoU of 0.79. In segmented cancer regions from whole slide images, the digital TPS achieved an accuracy of 75% against the manual TPS scores from the pathology report. Major reasons for algorithmic inaccuracies include the inclusion of immune cells in cancer outlines and poor nuclear segmentation of cancer cells. Our transparent and stepwise approach and performance metrics can be applied to any IHC assay to provide pathologists with important insights on when to apply and how to evaluate commercial automated IHC scoring systems.

Automated Gleason grading is an important preliminary step for quantitative histopathological feature extraction. Different from the traditional task of classifying small pre-selected homogeneous regions, semantic segmentation provides pixel-wise Gleason predictions across an entire slide. Deep learning-based segmentation models can automatically learn visual semantics from data, which alleviates the need for feature engineering. However, performance of deep learning models is limited by the scarcity of large-scale fully annotated datasets, which can be both expensive and time-consuming to create. One way to address this problem is to leverage external weakly labeled datasets to augment models trained on the limited data. In this paper, we developed an expectation maximization-based approach constrained by an approximated prior distribution in order to extract useful representations from a large number of weakly labeled images generated from low-magnification annotations. This method was utilized to improve the performance of a model trained on a limited fully annotated dataset. Our semi-supervised approach trained with 135 fully annotated and 1800 weakly annotated tiles achieved a mean Jaccard Index of 49.5% on an independent test set, which was 14% higher than the initial model trained only on the fully annotated dataset.

Extracellular vesicles (EVs) contain a wide range of RNA types with a reported prevalence of non-coding RNA. To date a comprehensive characterization of the protein coding transcripts in EVs is still lacking. We performed RNA-Sequencing (RNA-Seq) of 2 EV populations and identified a small fraction of transcripts that were expressed at significantly different levels in large oncosomes and exosomes, suggesting they may mediate specialized functions. However, these 2 EV populations exhibited a common mRNA signature that, in comparison to their donor cells, was significantly enriched in mRNAs encoding E2F transcriptional targets and histone proteins. These mRNAs are primarily expressed in the S-phase of the cell cycle, suggesting that they may be packaged into EVs during S-phase. In silico analysis using subcellular compartment transcriptome data from the ENCODE cell line compendium revealed that EV mRNAs originate from a cytoplasmic RNA pool. The EV signature was independently identified in plasma of patients with breast cancer by RNA-Seq. Furthermore, several transcripts differentially expressed in EVs from patients versus controls mirrored differential expression between normal and breast cancer tissues. Altogether, this largest high-throughput profiling of EV mRNA demonstrates that EVs carry tumor-specific alterations and can be interrogated as a source of cancer-derived cargo.

Background

Radiotherapy impacts the local immune response to cancers. Prostate Stereotactic Body Radiotherapy (SBRT) is a highly focused method to deliver radiotherapy often used to treat prostate cancer. This is the first direct comparison of immune cells within prostate cancers before and after SBRT in patients.

Methods

Prostate cancers before and 2 weeks after SBRT are interrogated by multiplex immune fluorescence targeting various T cells and macrophages markers and analyzed by cell and pixel density, as part of a clinical trial of SBRT neoadjuvant to radical prostatectomy.

Results

Two weeks after SBRT, CD68, and CD163 macrophages are significantly increased while CD8 T cells are decreased. SBRT markedly alters the immune environment within prostate cancers.

The CRISPR-based technology has revolutionized genome editing in recent years. This technique allows for gene knockout and evaluation of function in cell lines in a manner that is far easier and more accessible than anything previously available. Unfortunately, the ability to extend these studies to in vivo syngeneic murine cell line implantation is limited by an immune response against cells transduced to stably express Cas9. In this study, we demonstrate that a non-integrating lentiviral vector approach can overcome this immune rejection and allow for the growth of transduced cells in an immunocompetent host. This technique enables the establishment of a von Hippel-Lindau (VHL) gene knockout RENCA cell line in BALB/c mice, generating an improved model of immunocompetent, metastatic renal cell carcinoma (RCC).

Molecular and genomic analysis of microscopic quantities of tumor from formalin-fixed, paraffin-embedded biopsy specimens has many unique challenges. Herein, we evaluated the feasibility of obtaining transcriptome-wide RNA expression to measure prognostic classifiers in diagnostic prostate needle core biopsy specimens. One-hundred fifty-eight samples from diagnostic needle core biopsy specimens (BX) and radical prostatectomies (RPs) were collected from 33 patients at three hospitals; each patient provided up to six tumor and benign samples. Genome-wide transcriptomic profiles were generated using Affymetrix Human Exon arrays for comparison of gene expression alterations and prognostic signatures between the BX and RP samples. A sufficient amount of RNA (>100 ng) was obtained from all RP specimens (n = 77) and from 72 of 81 of BX specimens. Of transcriptomic features detected in RP, 95% were detectable in BX tissues and demonstrated a high correlation (r = 0.96). Likewise, an expression signature pattern validated on RPs (Decipher prognostic test) showed correlation between BX and RP (r = 0.70). Of matched BX and RP pairs, 25% showed discordant molecular subtypes. Genome-wide exon arrays yielded data of comparable quality from biopsy and RP tissues. The high concordance of tumor-associated gene expression changes between BX and RP samples provides evidence for the adequate performance of the assay platform with samples from prostate needle biopsy specimens with limited tumor volume.

Background

Benign human prostate tubule-initiating cells (TIC) and aggressive prostate cancer display common traits, including tolerance of low androgen levels, resistance to apoptosis, and microenvironment interactions that drive epithelial budding and outgrowth. TIC can be distinguished from epithelial and stromal cells that comprise prostate tissue via cell sorting based upon Epcam, CD44, and CD49f antigenic profiles. Fetal prostate epithelial cells (FC) possess a similar antigenic profile to adult TIC and are capable of inducing tubule formation. To identify the TIC niche in human prostate tissue, differential keratin (KRT) expression was evaluated.

Results

Gene expression data generated from Affymetrix Gene Chip human U133 Plus 2.0 array of sorted adult and fetal epithelial cells revealed KRT13 to be significantly enriched in FC and TIC compared to basal cells (BC) and luminal cells (LC) (p<0.001). Enriched KRT13 expression was confirmed by RT-PCR and cytospin immunostaining. Immunohistochemical analysis of KRT13 expression revealed rare KRT13+ epithelia throughout prostatic ducts/acini in adult tissue specimens and differentiated tubules in 24-week recombinant grafts, In contrast, abundant KRT13 expression was observed in developing ducts/acini in fetal prostate and cord-like structures composing 8-week recombinant grafts. Immunostaining of a prostate tissue microarray revealed KRT13+ tumor foci in approximately 9% of cases, and this subset displayed significantly shorter time to recurrence (p = 0.031), metastases (p = 0.032), and decreased overall survival (p = 0.004). Diagnostic prostate needle biopsies (PNBX) from untreated patients with concurrent bone metastases (clinical stage M1) displayed KRT13+ tumor foci, as did bone metastatic foci.

Conclusions

The expression profile of KRT13 in benign fetal and adult prostate tissue and in recombinant grafts, as well as the frequency of KRT13 expression in primary and metastatic prostate cancer indicates that it may be a marker of a stem/progenitor-like cell state that is co-opted in aggressive tumor cells. KRT13 is enriched in benign stem-like cells that display androgen-resistance, apoptosis-resistance, and branching morphogenesis properties. Collectively our data demonstrate that KRT13 expression is associated with poor prognosis at multiple stages of disease progression and may represent an important biomarker of adverse outcome in patients with prostate cancer.