Precision needle-punch tumor enrichment from paraffin blocks improves the detection of clinically actionable genomic alterations and biomarkers.

Background: While many molecular assays can detect mutations at low tumor purity and variant allele frequencies, complex biomarkers such as tumor mutational burden (TMB), microsatellite instability (MSI), and genomic loss of heterozygosity (gLOH) require higher tumor purity for accurate measurement. Scalable, quality-controlled, tissue-conserving methods to increase tumor nuclei percentage (TN%) from tumor specimens are needed for complex biomarkers and hence necessary to maximize patient matching to approved therapies or clinical trial enrollment. We evaluated the clinical utility and performance of precision needle-punch enrichment (NPE) compared with traditional razor blade macroenrichment of tumor specimens on molecular testing success. Methods: Pathologist-directed NPE was performed manually on formalin-fixed, paraffin embedded (FFPE) blocks. Quality control of target capture region and quantity of residual tumor in each tissue block was determined via a post-enrichment histologic slide recut. Resultant tumor purity and biomarker status were determined by the computational analysis pipeline component of the FDA-approved next-generation sequencing (NGS) assay, FoundationOne®CDx. Following NPE implementation for real-world clinical samples, assay performance and biomarker (MSI, TMB, gLOH) detection were analyzed. Results: In real-world clinical samples, enrichment rate via NPE was increased to ~50% over a 2.5-year period, exceeding the prior use of razor blade macro-enrichment (<30% of cases) prior to NPE implementation due to proven efficacy in generating high quality molecular results from marginal samples and the ease of use for both pathologist and histotechnologists. NPE was associated with lower test failures, higher computational tumor purity, and higher rates of successful TMB, MSI and gLOH determination when stratified by pre-enriched (incipient) tumor nuclei percentage. In addition, challenging cases in which tumor content was initially insufficient for testing were salvaged for analysis of biomarker status, gene amplification/deletion, and confident mutant or wild-type gene status determination. Conclusions: Pathologist-directed precision enrichment from tissue blocks (aka NPE) increases tumor purity, and consequently, yields a greater number of successful tests and complex biomarker determinations. Moreover, this process is rapid, safe, inexpensive, scalable, and conserves patient surgical pathology material. NPE may constitute best practice with respect to enriching tumor cells from low-purity specimens for biomarker detection in molecular laboratories.

Liquid biopsy-based circulating tumour (ct)DNA analysis of a spectrum of myeloid and lymphoid malignancies yields clinically actionable results.

AIMS: Liquid biopsy (LBx)-based next-generation sequencing (NGS) of circulating tumour DNA (ctDNA) can facilitate molecular profiling of haematopoietic neoplasms (HNs), particularly when tissue-based NGS is infeasible. METHODS AND RESULTS: We studied HN LBx samples tested with FoundationOne Liquid CDx, FoundationOne Liquid, or FoundationACT between July 2016 and March 2022. We identified 271 samples: 89 non-Hodgkin lymphoma (NHL), 43 plasma-cell neoplasm (PCN), 41 histiocytoses, 27 myelodysplastic syndrome (MDS), 25 diffuse large B-cell lymphoma (DLBCL), 22 myeloproliferative neoplasm (MPN), 14 Hodgkin lymphoma (HL), and 10 acute myeloid leukaemia (AML). Among 73.4% with detectable pathogenic alterations, median maximum somatic allele frequency (MSAF) was 16.6%, with AML (36.2%), MDS (19.7%), and MPN (44.5%) having higher MSAFs than DLBCL (3.9%), NHL (8.4%), HL (1.5%), PCN (2.8%), and histiocytoses (1.8%) (P = 0.001). LBx detected characteristic alterations across HNs, including in TP53, KRAS, MYD88, and BTK in NHLs; TP53, KRAS, NRAS, and BRAF in PCNs; IGH in DLBCL; TP53, ATM, and PDCD1LG2 in HL; BRAF and MAP2K1 in histiocytoses; TP53, SF3B1, DNMT3A, TET2, and ASXL1 in MDS; JAK2 in MPNs; and FLT3, IDH2, and NPM1 in AML. Among 24 samples, the positive percent agreement by LBx was 75.7% for variants present in paired buffy coat, marrow, or tissues. Also, 75.0% of pairs exhibited alterations only present on LBx. These were predominantly subclonal (clonal fraction of 3.8%), reflecting the analytical sensitivity of LBx. CONCLUSION: These data demonstrate that LBx can detect relevant genomic alterations across HNs, including at low clonal fractions, suggesting a potential clinical utility for identifying residual or emerging therapy-resistant clones that may be undetectable in site-specific tissue biopsies.

Methylthioadenosine Phosphorylase Genomic Loss in Advanced Gastrointestinal Cancers.

BACKGROUND: One of the most common sporadic homozygous deletions in cancers is 9p21 loss, which includes the genes methylthioadenosine phosphorylase (MTAP), CDKN2A, and CDKN2B, and has been correlated with worsened outcomes and immunotherapy resistance. MTAP-loss is a developing drug target through synthetic lethality with MAT2A and PMRT5 inhibitors. The purpose of this study is to investigate the prevalence and genomic landscape of MTAP-loss in advanced gastrointestinal (GI) tumors and investigate its role as a prognostic biomarker. MATERIALS AND METHODS: We performed next-generation sequencing and comparative genomic and clinical analysis on an extensive cohort of 64 860 tumors comprising 5 GI cancers. We compared the clinical outcomes of patients with GI cancer harboring MTAP-loss and MTAP-intact tumors in a retrospective study. RESULTS: The prevalence of MTAP-loss in GI cancers is 8.30%. MTAP-loss was most prevalent in pancreatic ductal adenocarcinoma (PDAC) at 21.7% and least in colorectal carcinoma (CRC) at 1.1%. MTAP-loss tumors were more prevalent in East Asian patients with PDAC (4.4% vs 3.2%, P = .005) or intrahepatic cholangiocarcinoma (IHCC; 6.4% vs 4.3%, P = .036). Significant differences in the prevalence of potentially targetable genomic alterations (ATM, BRAF, BRCA2, ERBB2, IDH1, PIK3CA, and PTEN) were observed in MTAP-loss tumors and varied according to tumor type. MTAP-loss PDAC, IHCC, and CRC had a lower prevalence of microsatellite instability or elevated tumor mutational burden. Positive PD-L1 tumor cell expression was less frequent among MTAP-loss versus MTAP-intact IHCC tumors (23.2% vs 31.2%, P = .017). CONCLUSION: In GI cancers, MTAP-loss occurs as part of 9p21 loss and has an overall prevalence of 8%. MTAP-loss occurs in 22% of PDAC, 15% of IHCC, 8.7% of gastroesophageal adenocarcinoma, 2.4% of hepatocellular carcinoma, and 1.1% of CRC and is not mutually exclusive with other targetable mutations.

Trait-Based Paleontological Niche Prediction Recovers Extinct Ecological Breadth of the Earliest Specialized Ant Predators.

AbstractPaleoecological estimation is fundamental to the reconstruction of evolutionary and environmental histories. The ant fossil record preserves a range of species in three-dimensional fidelity and chronicles faunal turnover across the Cretaceous and Cenozoic; taxonomically rich and ecologically diverse, ants are an exemplar system to test new methods of paleoecological estimation in evaluating hypotheses. We apply a broad extant ecomorphological dataset to evaluate random forest machine learning classification in predicting the total ecological breadth of extinct and enigmatic hell ants. In contrast to previous hypotheses of extinction-prone arboreality, we find that hell ants were primarily leaf litter or ground-nesting and foraging predators, and by comparing ecospace occupations of hell ants and their extant analogs, we recover a signature of ecomorphological turnover across temporally and phylogenetically distinct lineages on opposing sides of the Cretaceous-Paleogene boundary. This paleoecological predictive framework is applicable across lineages and may provide new avenues for testing hypotheses over deep time.

Genomic landscape of clinically advanced KRAS wild-type pancreatic ductal adenocarcinoma.

Introduction: KRAS mutation is a common occurrence in Pancreatic Ductal Adenocarcinoma (PDA) and is a driver mutation for disease development and progression. KRAS wild-type PDA may constitute a distinct molecular and clinical subtype. We used the Foundation one data to analyze the difference in Genomic Alterations (GAs) that occur in KRAS mutated and wild-type PDA. Methods: Comprehensive genomic profiling (CGP) data, tumor mutational burden (TMB), microsatellite instability (MSI) and PD-L1 by Immunohistochemistry (IHC) were analyzed. Results and discussion: Our cohort had 9444 cases of advanced PDA. 8723 (92.37%) patients had KRAS mutation. 721 (7.63%) patients were KRAS wild-type. Among potentially targetable mutations, GAs more common in KRAS wild-type included ERBB2 (mutated vs wild-type: 1.7% vs 6.8%, p <0.0001), BRAF (mutated vs wild-type: 0.5% vs 17.9%, p <0.0001), PIK3CA (mutated vs wild-type: 2.3% vs 6.5%, p <0.001), FGFR2 (mutated vs wild-type: 0.1% vs 4.4%, p <0.0001), ATM (mutated vs wild-type: 3.6% vs 6.8%, p <0.0001). On analyzing untargetable GAs, the KRAS mutated group had a significantly higher percentage of TP53 (mutated vs wild-type: 80.2% vs 47.6%, p <0.0001), CDKN2A (mutated vs wild-type: 56.2% vs 34.4%, p <0.0001), CDKN2B (mutated vs wild-type: 28.9% vs 23%, p =0.007), SMAD4 (mutated vs wild-type: 26.8% vs 15.7%, p <0.0001) and MTAP (mutated vs wild-type: 21.7% vs 18%, p =0.02). ARID1A (mutated vs wild-type: 7.7% vs 13.6%, p <0.0001 and RB1(mutated vs wild-type: 2% vs 4%, p =0.01) were more prevalent in the wild-type subgroup. Mean TMB was higher in the KRAS wild-type subgroup (mutated vs wild-type: 2.3 vs 3.6, p <0.0001). High TMB, defined as TMB > 10 mut/mB (mutated vs wild-type: 1% vs 6.3%, p <0.0001) and very-high TMB, defined as TMB >20 mut/mB (mutated vs wild-type: 0.5% vs 2.4%, p <0.0001) favored the wild-type. PD-L1 high expression was similar between the 2 groups (mutated vs wild-type: 5.7% vs 6%,). GA associated with immune checkpoint inhibitors (ICPIs) response including PBRM1 (mutated vs wild-type: 0.7% vs 3.2%, p <0.0001) and MDM2 (mutated vs wild-type: 1.3% vs 4.4%, p <0.0001) were more likely to be seen in KRAS wild-type PDA.

Novel SOX10 indel mutations drive schwannomas through impaired transactivation of myelination gene programs.

BACKGROUND: Schwannomas are common peripheral nerve sheath tumors that can cause severe morbidity given their stereotypic intracranial and paraspinal locations. Similar to many solid tumors, schwannomas and other nerve sheath tumors are primarily thought to arise due to aberrant hyperactivation of the RAS growth factor signaling pathway. Here, we sought to further define the molecular pathogenesis of schwannomas. METHODS: We performed comprehensive genomic profiling on a cohort of 96 human schwannomas, as well as DNA methylation profiling on a subset. Functional studies including RNA sequencing, chromatin immunoprecipitation-DNA sequencing, electrophoretic mobility shift assay, and luciferase reporter assays were performed in a fetal glial cell model following transduction with wildtype and tumor-derived mutant isoforms of SOX10. RESULTS: We identified that nearly one-third of sporadic schwannomas lack alterations in known nerve sheath tumor genes and instead harbor novel recurrent in-frame insertion/deletion mutations in SOX10, which encodes a transcription factor responsible for controlling Schwann cell differentiation and myelination. SOX10 indel mutations were highly enriched in schwannomas arising from nonvestibular cranial nerves (eg facial, trigeminal, vagus) and were absent from vestibular nerve schwannomas driven by NF2 mutation. Functional studies revealed these SOX10 indel mutations have retained DNA binding capacity but impaired transactivation of glial differentiation and myelination gene programs. CONCLUSIONS: We thus speculate that SOX10 indel mutations drive a unique subtype of schwannomas by impeding proper differentiation of immature Schwann cells.

Predicting EGFR mutational status from pathology images using a real-world dataset.

Treatment of non-small cell lung cancer is increasingly biomarker driven with multiple genomic alterations, including those in the epidermal growth factor receptor (EGFR) gene, that benefit from targeted therapies. We developed a set of algorithms to assess EGFR status and morphology using a real-world advanced lung adenocarcinoma cohort of 2099 patients with hematoxylin and eosin (H&E) images exhibiting high morphological diversity and low tumor content relative to public datasets. The best performing EGFR algorithm was attention-based and achieved an area under the curve (AUC) of 0.870, a negative predictive value (NPV) of 0.954 and a positive predictive value (PPV) of 0.410 in a validation cohort reflecting the 15% prevalence of EGFR mutations in lung adenocarcinoma. The attention model outperformed a heuristic-based model focused exclusively on tumor regions, and we show that although the attention model also extracts signal primarily from tumor morphology, it extracts additional signal from non-tumor tissue regions. Further analysis of high-attention regions by pathologists showed associations of predicted EGFR negativity with solid growth patterns and higher peritumoral immune presence. This algorithm highlights the potential of deep learning tools to provide instantaneous rule-out screening for biomarker alterations and may help prioritize the use of scarce tissue for biomarker testing.

Method of Tissue Acquisition Affects Success of Comprehensive Genomic Profiling in Lung Cancer.

CONTEXT.­: Multiple procedural techniques can be used to obtain tissue to create a formalin-fixed, paraffin-embedded specimen for comprehensive genomic profiling (CGP) in lung cancer. The literature is mixed on whether the procedure affects CGP success. OBJECTIVE.­: To examine whether biopsy procedure affects lung cancer CGP success. DESIGN.­: This was a cross-sectional study of all patients with lung cancer whose specimens were submitted for CGP between January and February 2020. Multiple quality control metrics were used to determine whether cases were successfully profiled. RESULTS.­: In all, 3312 samples were identified. Overall, 67.5% (2236 of 3312) of samples were obtained from biopsies, 13.0% (432 of 3312) from fine-needle aspirations (FNAs), 9.7% (321 of 3312) from resections, 5.3% (174 of 3312) from fluid cytology cell blocks, and 4.5% (149 of 3312) from bone biopsies. Overall, 70.1% (2321 of 3312) of cases passed CGP, 15.4% (510 of 3312) of cases were released as qualified reports, and 14.5% (481 of 3312) of cases failed CGP. Resection samples were the most likely to be successfully sequenced, failing in only 2.8% (9 of 321) of instances, while fluid cytology specimens were the least likely, failing in 23.0% (40 of 174) of instances. Biopsy (14.5% [324 of 2236]), FNA (18.5% [80 of 432]), and bone biopsy (18.8% [28 of 149]) specimens failed at intermediate frequencies. On multivariate logistic regression analysis of CGP success on specimen type, fluid cytology (odds ratio [OR], 0.08; 95% CI, 0.03-0.19), biopsy (OR, 0.25; 95% CI, 0.11-0.52), FNA (OR, 0.14; 95% CI, 0.06-0.32), and bone biopsy (OR, 0.07; 95% CI, 0.03-0.17) specimens had decreased odds of CGP success relative to resection samples. Among patients with successfully sequenced samples, 48.0% were eligible for at least 1 therapy, based on a companion diagnostic or National Comprehensive Cancer Network biomarker. CONCLUSIONS.­: The method of tissue acquisition was an important preanalytic factor that determined whether a sample would be successfully sequenced and whether a clinically actionable genomic alteration would be detected.

Activating IGF1R hotspot non-frameshift insertions define a novel, potentially targetable molecular subtype of adenoid cystic carcinoma.

Activation of the tyrosine kinase receptor IGF1R is targetable with existing tyrosine kinase inhibitors (TKIs) and monoclonal antibodies, but mutations in IGF1R have not been systematically characterized. Pan-cancer analysis of 326,911 tumors identified two distinct, activating non-frameshift insertion hotspots in IGF1R, which were significantly enriched in adenoid cystic carcinomas (ACCs). IGF1R alterations from 326,911 subjects were analyzed by variant effect prediction class, position within the gene, and cancer type. 6502 (2.0%) samples harbored one or more alterations in IGF1R. Two regions were enriched for non-frameshift insertions: codons 663-666 at the hinge region of the fibronectin type 3 domain and codons 1034-1049 in the tyrosine kinase domain. Hotspot insertions were highly enriched in ACCs (27.3-fold higher than in the remainder of the pan-cancer dataset; P = 2.3 × 10-17). Among salivary gland tumors, IGF1R hotspot insertions were entirely specific to ACCs. IGF1R alterations were most often mutually exclusive with other ACC drivers (9/15, 60%). Tumors with non-frameshift hotspot IGF1R insertions represent a novel, potentially targetable subtype of ACC. Additional studies are needed to determine whether these patients respond to existing IGF1R inhibitors.

Comprehensive Molecular Profiling of Oncocytic Salivary Gland Malignancies.

Oncocytic histologic features can be seen in a variety of salivary gland carcinomas. We performed a comprehensive molecular profiling of 15 salivary gland malignancies with oncocytic features (diagnosed as oncocytic carcinoma, carcinoma NOS with oncocytic features, or salivary duct carcinoma with oncocytic features). We reveal multiple novel molecular alterations that have not been previously described in other salivary gland malignancies, including, but not limited to, KEL amplification (13.3%, 2/15), PARP1 amplification (13.3%, 2/15), and EPHB4 amplification (13.3%, 2/15). Alterations in KMT2C (13.3%, 2/15), ERBB3 (13.3%, 2/15), CTNNA1 (13.3%, 2/15), and SMAD4 (20%, 3/15) were also found in this series and have been reported in other salivary gland malignancies. Alterations that have been reported in salivary duct carcinoma were also identified, including TP53 (40%, 6/15) , ERBB2 mutations (13.3%, 2/15) , ERBB2 amplification (13.3%, 2/15), PIK3CA (26.7%, 4/15) , PTEN (20%, 3/15), BRCA2 (20%, 3/15), BRAF (20%, 3/15), CDKN2A/B (20%, 3/15), CDH1 (13.3%, 2/15), and HRAS (13.3%, 2/15). Oncocytic salivary gland malignancies are a molecularly heterogenous group of tumors with partial overlap with salivary duct carcinoma subtypes.

Variable Genomic Landscapes of Advanced Melanomas with Heavy Pigmentation.

BACKGROUND: In the current study, we examined the real-world prevalence of highly pigmented advanced melanomas (HPMel) and the clinicopathologic, genomic, and ICPI biomarker signatures of this class of tumors. MATERIALS AND METHODS: Our case archive of clinical melanoma samples for which the ordering physician requested testing for both PD-L1 immunohistochemistry (IHC) and comprehensive genomic profiling (CGP) was screened for HPMel cases, as well as for non-pigmented or lightly pigmented advanced melanoma cases (LPMel). RESULTS: Of the 1268 consecutive melanoma biopsies in our archive that had been submitted for PD-L1 IHC, 13.0% (165/1268) were HPMel and 87.0% (1103/1268) were LPMel. In the HPMel cohort, we saw a significantly lower tumor mutational burden (TMB, median 8.8 mutations/Mb) than in the LPMel group (11.4 mut/Mb), although there was substantial overlap. In examining characteristic secondary genomic alterations (GA), we found that the frequencies of GA in TERTp, CDKN2A, TP53, and PTEN were significantly lower in the HPMel cases than in LPMel. A higher rate of GA in CTNNB1, APC, PRKAR1A, and KIT was identified in the HPMel cohort compared with LPMel. CONCLUSIONS: In this study, we quantified the failure rates of melanoma samples for PD-L1 testing due to high melanin pigmentation and showed that CGP can be used in these patients to identify biomarkers that can guide treatment decisions for HPMel patients. Using this practical clinical definition for tumor pigmentation, our results indicate that HPMel are frequent at 13% of melanoma samples, and in general appear molecularly less developed, with a lower TMB and less frequent secondary GA of melanoma progression.

Molecular analysis of endometrial serous carcinoma reveals distinct clinicopathologic and genomic subgroups.

OBJECTIVES: Endometrial serous carcinoma (EMSC) is an aggressive variant of uterine cancer with limited therapeutic options. We sought to define distinct clinicopathologic and genomic EMSC subgroups. METHODS: We retrospectively analyzed 2159 EMSC and 2346 endometrioid-type endometrial carcinomas (EEC) tissue specimens that had undergone comprehensive genomic profiling (CGP) via the FoundationOne CDx assay during routine clinical care. High tumor mutational burden (TMB) was defined as ≥10mut/Mb using the FDA-approved CDx cutoff for pembrolizumab. Microsatellite instability (MSI) was determined on 95 loci. Evidence of homologous recombination deficiency (HRD) was determined via genomic loss of heterozygosity (gLOH), a validated HRD detection method for predicting PARP inhibitor effectiveness in ovarian carcinoma. High gLOH was defined as ≥16%. RESULTS: A genomic analysis of 2159 EMSCs revealed a predominance of TP53 mutations, microsatellite stability, low tumor mutational burden (TMB), and recurrent alterations of PIK3CA, PPP2R1A, ERBB2, CCNE1, FBXW7 and MYC. Evidence of HRD via high gLOH was identified in 22% of EMSCs. BRCA1 and BRCA2 alterations, as well as unique SET (solid, pseudo-endometrioid, and transitional cell-like) variant morphology, were enriched in HRD-EMSC. There was an increased frequency of CCNE1 amplification, a lower prevalence of PIK3CA and PPP2R1A alterations, and no differences in HRD, MSI or TMB biomarker frequencies in patients of predicted African ancestry. EMSC exhibited distinct gene mutation frequencies and MSI, TMB and gLOH biomarker signatures compared to a cohort 2346 EEC. CONCLUSIONS: Molecularly defined subgroups provide a framework to test the susceptibility of EMSC to targeted therapies in specific genetic settings (e.g. HRD, PIK3CA, PPP2R1A, ERBB2, MYC, CCNE1).

Genomic profile of columnar cell variant of papillary thyroid carcinoma.

BACKGROUND AND AIMS: Columnar cell variant (CCV) is a rare papillary thyroid carcinoma subtype. The majority of CCV occur in older patients and are large, invasive tumours that pursue an aggressive clinical course. Rare well-circumscribed CCV occur in younger female patients and are comparatively indolent. METHODS AND RESULTS: We retrospectively identified CCV with material available to perform targeted next-generation sequencing and correlated molecular results with clinicopathological features and outcome. Our cohort was comprised of nine CCV. Nearly all were aggressive tumours; however, one was predominantly well-circumscribed and arose in a thyroglossal duct cyst of a 26-year-old woman who had no evidence of disease at last follow-up. Seven (78%) cases demonstrated activating oncogenic driver alterations in BRAF, including BRAF V600E, an activating N486_P490del deletion, and BRAF-AGK fusions. Activating RAS mutations were seen in two (22%) cases. Additionally, three (33%) cases had TERT promoter mutations, four (44%) had loss of the tumour suppressor CDKN2A and one (11%) case had a loss of function TP53 mutation. Most cases (89%) also demonstrated copy number alterations, including recurrent gain of chromosome 1q (five cases) and losses of chromosome 9p (three cases) and 22q (four cases). The one case without secondary pathogenic mutations or copy number alterations was the tumour in the 26-year-old woman. CONCLUSIONS: We found that CCV is primarily a BRAF-driven tumour, with most also harbouring secondary oncogenic mutations and multiple chromosomal gains and losses. Moreover, our findings suggest that molecular analysis could potentially be used to help risk stratify CCV.

Use of Deep Learning to Develop and Analyze Computational Hematoxylin and Eosin Staining of Prostate Core Biopsy Images for Tumor Diagnosis.

Importance: Histopathological diagnoses of tumors from tissue biopsy after hematoxylin and eosin (H&E) dye staining is the criterion standard for oncological care, but H&E staining requires trained operators, dyes and reagents, and precious tissue samples that cannot be reused. Objectives: To use deep learning algorithms to develop models that perform accurate computational H&E staining of native nonstained prostate core biopsy images and to develop methods for interpretation of H&E staining deep learning models and analysis of computationally stained images by computer vision and clinical approaches. Design, Setting, and Participants: This cross-sectional study used hundreds of thousands of native nonstained RGB (red, green, and blue channel) whole slide image (WSI) patches of prostate core tissue biopsies obtained from excess tissue material from prostate core biopsies performed in the course of routine clinical care between January 7, 2014, and January 7, 2017, at Brigham and Women's Hospital, Boston, Massachusetts. Biopsies were registered with their H&E-stained versions. Conditional generative adversarial neural networks (cGANs) that automate conversion of native nonstained RGB WSI to computational H&E-stained images were then trained. Deidentified whole slide images of prostate core biopsy and medical record data were transferred to Massachusetts Institute of Technology, Cambridge, for computational research. Results were shared with physicians for clinical evaluations. Data were analyzed from July 2018 to February 2019. Main Outcomes and Measures: Methods for detailed computer vision image analytics, visualization of trained cGAN model outputs, and clinical evaluation of virtually stained images were developed. The main outcome was interpretable deep learning models and computational H&E-stained images that achieved high performance in these metrics. Results: Among 38 patients who provided samples, single core biopsy images were extracted from each whole slide, resulting in 102 individual nonstained and H&E dye-stained image pairs that were compared with matched computationally stained and unstained images. Calculations showed high similarities between computationally and H&E dye-stained images, with a mean (SD) structural similarity index (SSIM) of 0.902 (0.026), Pearson correlation coefficient (PCC) of 0.962 (0.096), and peak signal to noise ratio (PSNR) of 22.821 (1.232) dB. A second cGAN performed accurate computational destaining of H&E-stained images back to their native nonstained form, with a mean (SD) SSIM of 0.900 (0.030), PCC of 0.963 (0.011), and PSNR of 25.646 (1.943) dB compared with native nonstained images. A single blind prospective study computed approximately 95% pixel-by-pixel overlap among prostate tumor annotations provided by 5 board certified pathologists on computationally stained images, compared with those on H&E dye-stained images. This study also used the first visualization and explanation of neural network kernel activation maps during H&E staining and destaining of RGB images by cGANs. High similarities between kernel activation maps of computationally and H&E-stained images (mean-squared errors <0.0005) provide additional mathematical and mechanistic validation of the staining system. Conclusions and Relevance: These findings suggest that computational H&E staining of native unlabeled RGB images of prostate core biopsy could reproduce Gleason grade tumor signatures that were easily assessed and validated by clinicians. Methods for benchmarking, visualization, and clinical validation of deep learning models and virtually H&E-stained images communicated in this study have wide applications in clinical informatics and oncology research. Clinical researchers may use these systems for early indications of possible abnormalities in native nonstained tissue biopsies prior to histopathological workflows.

Clinically Relevant Prognostic Parameters in Differentiated Thyroid Carcinoma.

Although differentiated thyroid carcinomas typically pursue an indolent clinical course, it is important to identify the subset of tumors that are most likely to behave aggressively so that patients with these tumors are counseled and treated appropriately. Extent of disease is fundamental to the prognostication for differentiated thyroid carcinoma; however, there are additional histologic features of the tumor separate from extent of disease that have been shown to affect clinical course. This review will start with a discussion of aggressive variants of papillary thyroid carcinoma, move to the prognostic significance of vascular invasion in follicular thyroid carcinoma, and finish with a discussion of Hürthle cell carcinoma, with an emphasis on why it is not considered a subtype of follicular thyroid carcinoma in the 2017 WHO Classification of Tumors of Endocrine Organs.

Parvovirus B19 integration into human CD36+ erythroid progenitor cells.

The pathogenic autonomous human parvovirus B19 (B19V) productively infects erythroid progenitor cells (EPCs). Functional similarities between B19V nonstructural protein (NS1), a DNA binding endonuclease, and the Rep proteins of Adeno-Associated Virus (AAV) led us to hypothesize that NS1 may facilitate targeted nicking of the human genome and B19 vDNA integration. We adapted an integration capture sequencing protocol (IC-Seq) to screen B19V infected human CD36+ EPCs for viral integrants, and discovered 40,000 unique B19V integration events distributed throughout the human genome. Computational analysis of integration patterns revealed strong correlations with gene intronic regions, H3K9me3 sites, and the identification of 41 base pair consensus sequence with an octanucleotide core motif. The octanucleotide core has homology to a single region of B19V, adjacent to the P6 promoter TATA box. We present the first direct evidence that B19V infection of erythroid progenitor cells disrupts the human genome and facilitates viral DNA integration.

Adeno-associated virus type 2 preferentially integrates single genome copies with defined breakpoints.

BACKGROUND: Adeno-associated virus (AAV) serotype 2 prevalently infects humans and is the only described eukaryotic virus that integrates site-preferentially. In a recent high throughput study, the genome wide distribution of AAV-2 integrants was determined using Integrant Capture Sequencing (IC-Seq). Additional insight regarding the integration of AAV-2 into human genomic DNA could be gleaned by low-throughput sequencing of complete viral-chromosomal junctions. FINDINGS: In this study, 140 clones derived from Integrant-Capture Sequencing were sequenced. 100 met sequence inclusion criteria, and of these 39 contained validated junction sequences. These unique sequences were analyzed to investigate the structure and location of viral-chromosomal junctions. CONCLUSIONS: Overall the low-throughput analysis confirmed the genome wide distribution profile gathered through the IC-Seq analysis. We found no unidentifiable sequence inserted at AAV-2 chromosomal junctions. Assessing both left and right ends of the AAV genome, viral breakpoints predominantly occurred in one hairpin of the inverted terminal repeat and AAV genomes were preferentially integrated as single copies.

Highly divergent integration profile of adeno-associated virus serotype 5 revealed by high-throughput sequencing.

UNLABELLED: Adeno-associated virus serotype 5 (AAV-5) is a human parvovirus that infects a high percentage of the population. It is the most divergent AAV, the DNA sequence cleaved by the viral endonuclease is distinct from all other described serotypes and, uniquely, AAV-5 does not cross-complement the replication of other serotypes. In contrast to the well-characterized integration of AAV-2, no published studies have investigated the genomic integration of AAV-5. In this study, we analyzed more than 660,000 AAV-5 integration junctions using high-throughput integrant capture sequencing of infected human cells. The integration activity of AAV-5 was 99.7% distinct from AAV-2 and favored intronic sequences. Genome-wide integration was highly correlated with viral replication protein binding and endonuclease sites, and a 39-bp consensus integration motif was revealed that included these features. Algorithmic scanning identified 126 AAV-5 hot spots, the largest of which encompassed 3.3% of all integration events. The unique aspects of AAV-5 integration may provide novel tools for biotechnology and gene therapy. IMPORTANCE: Viral integration into the host genome is an important aspect of virus host cell biology. Genomic integration studies of the small single-stranded AAVs have largely focused on site preferential integration of AAV-2, which depends on the viral replication protein (Rep). We have now established the first genome wide integration profile of the highly divergent AAV-5 serotype. Using integrant capture sequencing, more than 600,000 AAV-5 integration junctions in human cells were analyzed. AAV-5 integration hot spots were 99.7% distinct from AAV-2. Integration favored intronic sequences, occurred on all chromosomes, and integration hot spot distribution was correlated with human genomic GAGC repeats and transcriptional activity. These features support expansion of AAV-5 based vectors for gene transfer considerations.

High-throughput sequencing reveals principles of adeno-associated virus serotype 2 integration.

Viral integrations are important in human biology, yet genome-wide integration profiles have not been determined for many viruses. Adeno-associated virus (AAV) infects most of the human population and is a prevalent gene therapy vector. AAV integrates into the human genome with preference for a single locus, termed AAVS1. However, the genome-wide integration of AAV has not been defined, and the principles underlying this recombination remain unclear. Using a novel high-throughput approach, integrant capture sequencing, nearly 12 million AAV junctions were recovered from a human cell line, providing five orders of magnitude more data than were previously available. Forty-five percent of integrations occurred near AAVS1, and several thousand novel integration hotspots were identified computationally. Most of these occurred in genes, with dozens of hotspots targeting known oncogenes. Viral replication protein binding sites (RBS) and transcriptional activity were major factors favoring integration. In a first for eukaryotic viruses, the data reveal a unique asymmetric integration profile with distinctive directional orientation of viral genomes. These studies provide a new understanding of AAV integration biology through the use of unbiased high-throughput data acquisition and bioinformatics.

Expansion and maintenance of human embryonic stem cell-derived endothelial cells by TGFbeta inhibition is Id1 dependent.

Previous efforts to differentiate human embryonic stem cells (hESCs) into endothelial cells have not achieved sustained expansion and stability of vascular cells. To define vasculogenic developmental pathways and enhance differentiation, we used an endothelial cell-specific VE-cadherin promoter driving green fluorescent protein (GFP) (hVPr-GFP) to screen for factors that promote vascular commitment. In phase 1 of our method, inhibition of transforming growth factor (TGF)beta at day 7 of differentiation increases hVPr-GFP(+) cells by tenfold. In phase 2, TGFbeta inhibition maintains the proliferation and vascular identity of purified endothelial cells, resulting in a net 36-fold expansion of endothelial cells in homogenous monolayers, which exhibited a transcriptional profile of Id1(high)VEGFR2(high)VE-cadherin(+) ephrinB2(+). Using an Id1-YFP hESC reporter line, we showed that TGFbeta inhibition sustains Id1 expression in hESC-derived endothelial cells and that Id1 is required for increased proliferation and preservation of endothelial cell commitment. Our approach provides a serum-free method for differentiation and long-term maintenance of hESC-derived endothelial cells at a scale relevant to clinical application.