Deep learning image analysis quantifies tumor heterogeneity and identifies microsatellite instability in colon cancer.

BACKGROUND AND OBJECTIVES: Deep learning utilizing convolutional neural networks (CNNs) applied to hematoxylin & eosin (H&E)-stained slides numerically encodes histomorphological tumor features. Tumor heterogeneity is an emerging biomarker in colon cancer that is, captured by these features, whereas microsatellite instability (MSI) is an established biomarker traditionally assessed by immunohistochemistry or polymerase chain reaction. METHODS: H&E-stained slides from The Cancer Genome Atlas (TCGA) colon cohort are passed through the CNN. Resulting imaging features are used to cluster morphologically similar slide regions. Tile-level pairwise similarities are calculated and used to generate a tumor heterogeneity score (THS). Patient-level THS is then correlated with TCGA-reported biomarkers, including MSI-status. RESULTS: H&E-stained images from 313 patients generated 534 771 tiles. Deep learning automatically identified and annotated cells by type and clustered morphologically similar slide regions. MSI-high tumors demonstrated significantly higher THS than MSS/MSI-low (p < 0.001). THS was higher in MLH1-silent versus non-silent tumors (p < 0.001). The sequencing derived MSIsensor score also correlated with THS (r = 0.51, p < 0.0001). CONCLUSIONS: Deep learning provides spatially resolved visualization of imaging-derived biomarkers and automated quantification of tumor heterogeneity. Our novel THS correlates with MSI-status, indicating that with expanded training sets, translational tools could be developed that predict MSI-status using H&E-stained images alone.

Deconvolution microscopy: A platform for rapid on-site evaluation of fine needle aspiration specimens that enables recovery of the sample.

CONTEXT: Rapid on-site evaluation (ROSE) optimises the performance of cytology, but requires skilled handling, and smearing can make the material unavailable for some ancillary tests. There is a need to facilitate ROSE without sacrificing part of the sample. OBJECTIVE: We evaluated the image quality of inexpensive deconvolution fluorescence microscopy for optically sectioning non-smeared fine needle aspiration (FNA) tissue fragments. DESIGN: A portion of residual material from 14 FNA samples was stained for 3 min in Hoechst 33342 and Sypro™ Red to label DNA and protein respectively, transferred to an imaging chamber, and imaged at 200× or 400× magnification at 1 micron intervals using a GE DeltaVision inverted fluorescence microscope. A deconvolution algorithm was applied to remove out-of-plane signal, and the resulting images were inverted and pseudocoloured to resemble H&E sections. Five cytopathologists blindly diagnosed 2 to 4 representative image stacks per case (total 70 evaluations), and later compared them to conventional epifluorescent images. RESULTS: Accurate definitive diagnoses were rendered in 45 (64%) of 70 total evaluations; equivocal diagnoses (atypical or suspicious) were made in 21 (30%) of the 70. There were two false positive and two false negative "definite" diagnoses in three cases (4/70; 6%). Cytopathologists preferred deconvolved images compared to raw images (P < 0.01). The imaged fragments were recovered and prepared into a ThinPrep or cell block without discernible alteration. CONCLUSIONS: Deconvolution improves image quality of FNA fragments compared to epifluorescence, often allowing definitive diagnosis while enabling the ROSE material to be subsequently triaged.

Significance of Lymphovascular Space Invasion by the Sarcomatous Component in Uterine Carcinosarcoma.

OBJECTIVE: The aim of this study was to examine the significance of lymphovascular space invasion (LVSI) with a sarcomatous component on the tumor characteristics and clinical outcomes of women with uterine carcinosarcoma (UCS). METHODS: This was a secondary analysis of a prior multicenter retrospective study that examined women with stage I-IV UCS who underwent primary hysterectomy. Archived histopathology slides were reviewed and LVSI was scored as follows: LVSI with a carcinomatous component alone (LVSI-carcinoma; n = 375, 76.8%) or LVSI containing a sarcomatous component with or without a carcinomatous component (LVSI-sarcoma; n = 113, 23.2%). Qualitative metrics of LVSI were correlated to clinicopathological factors and survival outcome. RESULTS: Tumors in the LVSI-sarcoma group were more likely to have sarcoma dominance (82.1 vs. 26.4%) heterologous sarcomatous component (51.3 vs. 37.9%), low-grade carcinoma (42.5 vs. 22.4%), and large tumor size (81.0 vs. 70.2%) in the primary tumor site compared with tumors in the LVSI-carcinoma group (all p < 0.05). On multivariate analysis, LVSI-sarcoma was independently associated with decreased progression-free survival (5-year rates: 34.9 vs. 40.8%, adjusted hazard ratio [HR] 1.84, 95% confidence interval [CI] 1.36-2.50, p < 0.001), and cause-specific survival (5-year rates: 41.8 vs. 55.9%, adjusted HR 1.95, 95% CI 1.39-2.75, p < 0.001) compared with LVSI-carcinoma. Postoperative radiotherapy for women with LVSI-sarcoma had a higher reduction rate of recurrence/progression of disease (54% reduction, p = 0.04) compared with postoperative radiotherapy for women with LVSI-carcinoma (26% reduction, p = 0.08). CONCLUSION: In UCS, the presence of a sarcomatous component in LVSI is particularly prevalent when a tumor has sarcoma dominance. Our study suggests that LVSI containing a sarcomatous component may be a predictor of decreased survival for women with UCS.

Proposal for a Risk-Based Categorization of Uterine Carcinosarcoma.

PURPOSE: To propose a categorization model of uterine carcinosarcoma (UCS) based on tumor cell types (carcinoma and sarcoma) and sarcoma dominance. METHODS: This secondary analysis of a prior multicenter retrospective study examined 889 cases of UCS with available histologic evaluation. Based on survival outcome, cases were clustered into three groups: low-grade carcinoma with nondominant homologous sarcoma [type A, n = 96 (10.8%)], (1) low-grade carcinoma with heterologous sarcoma or any sarcoma dominance and (2) high-grade carcinoma with nondominant homologous sarcoma [type B, n = 412 (46.3%)], and high-grade carcinoma with heterologous sarcoma or any sarcoma dominance [type C, n = 381 (42.9%)]. Tumor characteristics and outcome were examined based on the categorization. RESULTS: Women in type C category were more likely to be older, obese, and Caucasian, whereas those in type A category were younger, less obese, Asian, and nulligravid (all P < 0.01). Type C tumors were more likely to have metastatic implants, large tumor size, lymphovascular space invasion with sarcoma cells, and higher lymph node ratio, whereas type A tumors were more likely to be early-stage disease and small (all P < 0.05). On multivariate analysis, tumor categorization was independently associated with progression-free survival (5-year rates: 70.1% for type A, 48.3% for type B, and 35.9% for type C, adjusted P < 0.01) and cause-specific survival (5-year rates: 82.8% for type A, 63.0% for type B, and 47.1% for type C, adjusted P < 0.01). CONCLUSION: Characteristic differences in clinicopathological factors and outcomes in UCS imply that different underlying etiologies and biological behaviors may be present, supporting a new classification system.

Survival outcome of stage I ovarian clear cell carcinoma with lympho-vascular space invasion.

BACKGROUND: The clinical impact of lympho-vascular space invasion (LVSI) in early-stage ovarian clear cell carcinoma (OCCC) is not well understood. Given the distinct tumor biology and survival patterns of OCCC, the significance of LVSI on survival outcome and treatment response was examined in OCCC. METHODS: A multicenter study was conducted to examine stage IA-IC3 OCCC cases that underwent primary surgical staging including lymphadenectomy. LVSI status was determined from archived histopathology slides, correlated with clinico-pathological results, chemotherapy patterns, and survival outcomes. RESULTS: LVSI was observed in 47 (20.3%) among 232 cases. In univariate analysis, LVSI was associated with older age (p=0.042), large tumor size (p=0.048), and stage IC (p=0.035). In survival analysis, LVSI was associated with decreased disease-free survival (DFS, 5-year rate, 70.6% versus 92.1%, p=0.0004) and overall survival (OS, 78.8% versus 93.3%, p=0.008) on univariate analysis. After controlling for age, tumor size, stage, and chemotherapy use, LVSI remained an independent prognostic factor for decreased survival outcomes (DFS, hazard ratio [HR] 4.35, 95% confidence interval [CI] 1.73-10.9, p=0.002; and OS, HR 4.73, 95%CI 1.60-14.0, p=0.015). Among 210 cases who received postoperative chemotherapy, while regimen type did not impact survival outcome regardless of LVSI status (DFS, p=0.63), the number of administered cycles showed a survival benefit towards ≥6cycles for patients with LVSI-positive tumors (DFS, p=0.009; and OS, p=0.016). CONCLUSION: LVSI is an important marker to predict survival outcome of stage I OCCC. Regardless of chemotherapy type, patients with stage I OCCC showing LVSI may benefit from receiving postoperative chemotherapy.

Estrogen receptor expression and increased risk of lymphovascular space invasion in high-grade serous ovarian carcinoma.

OBJECTIVE: Recent studies have demonstrated that lymphovascular space invasion (LVSI) is associated with increased risk of hematogenous and lymphatic metastasis and poor clinical outcome of women with epithelial ovarian cancer. Given the suspected role of estrogen in promoting ovarian cancer metastasis, we examined potential links between estrogen receptor and LVSI in high-grade serous ovarian carcinoma. METHODS: Tumoral expression of ER, PR, p53, MDR1, EGFR, HER2, DNA ploidy, and S-phase fraction was examined for 121 cases of stage I-IV high-grade serous ovarian carcinoma samples obtained at primary cytoreductive surgery. Biomarker expression was correlated to LVSI and survival outcomes. RESULTS: LVSI was observed in 101 (83.5%) of all cases. Immunohistochemistry of tested biomarkers showed ER (86.7%) to be the most commonly expressed followed by p53 (71.4%), HER2 (68.3%), EGFR (52.1%), MDR-1 (14.3%), and PR (8.9%). ER expression was positively correlated to PR expression (r=0.31, p=0.001). LVSI was only correlated with ER (odds ratio 6.27, 95%CI 1.93-20.4, p=0.002) but not with other biomarkers. In multivariate analysis, ER remained significantly associated with LVSI (p=0.039). LVSI remained a significant prognostic factor for decreased progression-free survival (HR 3.01, 95%CI 1.54-5.88, p=0.001) and overall survival (HR 2.69, 95%CI 1.18-6.23, p=0.021) while ER-expression did not remain as a significant variable in multivariate analysis. CONCLUSION: Our data demonstrated that estrogen receptor was positively correlated with LVSI that was an independent prognostic indicator of poor survival outcomes of high-grade serous ovarian carcinoma. This study emphasizes the importance of estrogen pathway in promoting lymphatic or vascular spread of high-grade serous ovarian carcinoma.

SAMPLER: unsupervised representations for rapid analysis of whole slide tissue images.

BACKGROUND: Deep learning has revolutionized digital pathology, allowing automatic analysis of hematoxylin and eosin (H&E) stained whole slide images (WSIs) for diverse tasks. WSIs are broken into smaller images called tiles, and a neural network encodes each tile. Many recent works use supervised attention-based models to aggregate tile-level features into a slide-level representation, which is then used for downstream analysis. Training supervised attention-based models is computationally intensive, architecture optimization of the attention module is non-trivial, and labeled data are not always available. Therefore, we developed an unsupervised and fast approach called SAMPLER to generate slide-level representations. METHODS: Slide-level representations of SAMPLER are generated by encoding the cumulative distribution functions of multiscale tile-level features. To assess effectiveness of SAMPLER, slide-level representations of breast carcinoma (BRCA), non-small cell lung carcinoma (NSCLC), and renal cell carcinoma (RCC) WSIs of The Cancer Genome Atlas (TCGA) were used to train separate classifiers distinguishing tumor subtypes in FFPE and frozen WSIs. In addition, BRCA and NSCLC classifiers were externally validated on frozen WSIs. Moreover, SAMPLER's attention maps identify regions of interest, which were evaluated by a pathologist. To determine time efficiency of SAMPLER, we compared runtime of SAMPLER with two attention-based models. SAMPLER concepts were used to improve the design of a context-aware multi-head attention model (context-MHA). FINDINGS: SAMPLER-based classifiers were comparable to state-of-the-art attention deep learning models to distinguish subtypes of BRCA (AUC = 0.911 ± 0.029), NSCLC (AUC = 0.940 ± 0.018), and RCC (AUC = 0.987 ± 0.006) on FFPE WSIs (internal test sets). However, training SAMLER-based classifiers was >100 times faster. SAMPLER models successfully distinguished tumor subtypes on both internal and external test sets of frozen WSIs. Histopathological review confirmed that SAMPLER-identified high attention tiles contained subtype-specific morphological features. The improved context-MHA distinguished subtypes of BRCA and RCC (BRCA-AUC = 0.921 ± 0.027, RCC-AUC = 0.988 ± 0.010) with increased accuracy on internal test FFPE WSIs. INTERPRETATION: Our unsupervised statistical approach is fast and effective for analyzing WSIs, with greatly improved scalability over attention-based deep learning methods. The high accuracy of SAMPLER-based classifiers and interpretable attention maps suggest that SAMPLER successfully encodes the distinct morphologies within WSIs and will be applicable to general histology image analysis problems. FUNDING: This study was supported by the National Cancer Institute (Grant No. R01CA230031 and P30CA034196).

Spatiotemporal profiling defines persistence and resistance dynamics during targeted treatment of melanoma.

Resistance of BRAF-mutant melanomas to targeted therapy arises from the ability of cells to enter a persister state, evade treatment with relative dormancy, and repopulate the tumor when reactivated. Using spatial transcriptomics in patient derived xenograft models, we capture clonal lineage evolution during treatment, finding the persister state to show increased oxidative phosphorylation, decreased proliferation, and increased invasive capacity, with central-to-peripheral gradients. Phylogenetic tracing identifies intrinsic- and acquired-resistance mechanisms (e.g. dual specific phosphatases, Reticulon-4, CDK2) and suggests specific temporal windows of potential therapeutic efficacy. Using deep learning to analyze histopathological slides, we find morphological features of specific cell states, demonstrating that juxtaposition of transcriptomics and histology data enables identification of phenotypically-distinct populations using imaging data alone. In summary, we define state change and lineage selection during melanoma treatment with spatiotemporal resolution, elucidating how choice and timing of therapeutic agents will impact the ability to eradicate resistant clones. Statement of Significance: Tumor evolution is accelerated by application of anti-cancer therapy, resulting in clonal expansions leading to dormancy and subsequently resistance, but the dynamics of this process are incompletely understood. Tracking clonal progression during treatment, we identify conserved, global transcriptional changes and local clone-clone and spatial patterns underlying the emergence of resistance.

A Pan-Cancer Patient-Derived Xenograft Histology Image Repository with Genomic and Pathologic Annotations Enables Deep Learning Analysis.

Patient-derived xenografts (PDX) model human intra- and intertumoral heterogeneity in the context of the intact tissue of immunocompromised mice. Histologic imaging via hematoxylin and eosin (H&E) staining is routinely performed on PDX samples, which could be harnessed for computational analysis. Prior studies of large clinical H&E image repositories have shown that deep learning analysis can identify intercellular and morphologic signals correlated with disease phenotype and therapeutic response. In this study, we developed an extensive, pan-cancer repository of >1,000 PDX and paired parental tumor H&E images. These images, curated from the PDX Development and Trial Centers Research Network Consortium, had a range of associated genomic and transcriptomic data, clinical metadata, pathologic assessments of cell composition, and, in several cases, detailed pathologic annotations of neoplastic, stromal, and necrotic regions. The amenability of these images to deep learning was highlighted through three applications: (i) development of a classifier for neoplastic, stromal, and necrotic regions; (ii) development of a predictor of xenograft-transplant lymphoproliferative disorder; and (iii) application of a published predictor of microsatellite instability. Together, this PDX Development and Trial Centers Research Network image repository provides a valuable resource for controlled digital pathology analysis, both for the evaluation of technical issues and for the development of computational image-based methods that make clinical predictions based on PDX treatment studies. Significance: A pan-cancer repository of >1,000 patient-derived xenograft hematoxylin and eosin-stained images will facilitate cancer biology investigations through histopathologic analysis and contributes important model system data that expand existing human histology repositories.

SAMPLER: Empirical distribution representations for rapid analysis of whole slide tissue images.

Deep learning has revolutionized digital pathology, allowing for automatic analysis of hematoxylin and eosin (H&E) stained whole slide images (WSIs) for diverse tasks. In such analyses, WSIs are typically broken into smaller images called tiles, and a neural network backbone encodes each tile in a feature space. Many recent works have applied attention based deep learning models to aggregate tile-level features into a slide-level representation, which is then used for slide-level prediction tasks. However, training attention models is computationally intensive, necessitating hyperparameter optimization and specialized training procedures. Here, we propose SAMPLER, a fully statistical approach to generate efficient and informative WSI representations by encoding the empirical cumulative distribution functions (CDFs) of multiscale tile features. We demonstrate that SAMPLER-based classifiers are as accurate or better than state-of-the-art fully deep learning attention models for classification tasks including distinction of: subtypes of breast carcinoma (BRCA: AUC=0.911 ± 0.029); subtypes of non-small cell lung carcinoma (NSCLC: AUC=0.940±0.018); and subtypes of renal cell carcinoma (RCC: AUC=0.987±0.006). A major advantage of the SAMPLER representation is that predictive models are >100X faster compared to attention models. Histopathological review confirms that SAMPLER-identified high attention tiles contain tumor morphological features specific to the tumor type, while low attention tiles contain fibrous stroma, blood, or tissue folding artifacts. We further apply SAMPLER concepts to improve the design of attention-based neural networks, yielding a context aware multi-head attention model with increased accuracy for subtype classification within BRCA and RCC (BRCA: AUC=0.921±0.027, and RCC: AUC=0.988±0.010). Finally, we provide theoretical results identifying sufficient conditions for which SAMPLER is optimal. SAMPLER is a fast and effective approach for analyzing WSIs, with greatly improved scalability over attention methods to benefit digital pathology analysis.

Deep learning features encode interpretable morphologies within histological images.

Convolutional neural networks (CNNs) are revolutionizing digital pathology by enabling machine learning-based classification of a variety of phenotypes from hematoxylin and eosin (H&E) whole slide images (WSIs), but the interpretation of CNNs remains difficult. Most studies have considered interpretability in a post hoc fashion, e.g. by presenting example regions with strongly predicted class labels. However, such an approach does not explain the biological features that contribute to correct predictions. To address this problem, here we investigate the interpretability of H&E-derived CNN features (the feature weights in the final layer of a transfer-learning-based architecture). While many studies have incorporated CNN features into predictive models, there has been little empirical study of their properties. We show such features can be construed as abstract morphological genes ("mones") with strong independent associations to biological phenotypes. Many mones are specific to individual cancer types, while others are found in multiple cancers especially from related tissue types. We also observe that mone-mone correlations are strong and robustly preserved across related cancers. Importantly, linear mone-based classifiers can very accurately separate 38 distinct classes (19 tumor types and their adjacent normals, AUC = [Formula: see text] for each class prediction), and linear classifiers are also highly effective for universal tumor detection (AUC = [Formula: see text]). This linearity provides evidence that individual mones or correlated mone clusters may be associated with interpretable histopathological features or other patient characteristics. In particular, the statistical similarity of mones to gene expression values allows integrative mone analysis via expression-based bioinformatics approaches. We observe strong correlations between individual mones and individual gene expression values, notably mones associated with collagen gene expression in ovarian cancer. Mone-expression comparisons also indicate that immunoglobulin expression can be identified using mones in colon adenocarcinoma and that immune activity can be identified across multiple cancer types, and we verify these findings by expert histopathological review. Our work demonstrates that mones provide a morphological H&E decomposition that can be effectively associated with diverse phenotypes, analogous to the interpretability of transcription via gene expression values. Our work also demonstrates mones can be interpreted without using a classifier as a proxy.

Genetically diverse mouse platform to xenograft cancer cells.

The lack of genetically diverse preclinical animal models in basic biology and efficacy testing has been cited as a potential cause of failure in clinical trials. We developed and characterized five diverse RAG1 null mouse strains as models that allow xenografts to grow. In these strains, we characterized the growth of breast cancer, leukemia and glioma cell lines. We found a wide range of growth characteristics that were far more dependent on strain than tumor type. For the breast cancer cell line, we characterized the spectrum of xenograft/tumor growth at structural, histological, cellular and molecular levels across each strain, and found that each strain captures unique structural components of the stroma. Furthermore, we showed that the increase in tumor-infiltrating myeloid CD45+ cells and the amount of circulating cytokine IL-6 and chemokine KC (also known as CXCL1) is associated with a higher tumor size in different strains. This resource is available to study established human xenografts, as well as difficult-to-xenograft tumors and growth of hematopoietic stems cells, and to decipher the role of myeloid cells in the development of spontaneous cancers.

Cytologic processing of ureteral microbiopsies is associated with higher sensitivity for detection of urothelial carcinoma compared with conventional biopsy processing.

INTRODUCTION: Because of the high rates of false-negative or nondiagnostic ureteral Piranha microbiopsies associated with low cellularity, we assessed the effect of processing these using cytology. MATERIALS AND METHODS: We included 2 groups of 44 consecutive microbiopsies processed from formalin as a standard surgical biopsy and 22 processed by cytology. All samples were from the ureter or renal pelvis or calyx. The cytology samples were collected in alcohol-based media and were prepared with a Cellient cell block only (n = 9) or with a Cellient cell block for the visible particles, together with ThinPrep, to capture the remaining desquamated cells (n = 13). RESULTS: Malignancy was diagnosed in 5 of 44 conventionally processed microbiopsies (11%) compared with 14 of 22 cytologically processed microbiopsies (64%; P < 0.001), including 1 case with invasion. Nineteen site-matched biopsies from 2 patients had undergone both cytologic and surgical processing, with 8 of 8 cytologically processed biopsies diagnosed as malignant. None of the 11 surgically processed biopsies from the same patients matched for site were diagnosed as malignant. Of the 11, 2 (18%) were suspicious for high-grade urothelial carcinoma and 6 (55%) were considered atypical. Increased sensitivity from cytologic processing appears related to increased cell recovery; large numbers of well-preserved urothelial cells were identified in the ThinPrep (range, 1000-25,000 cells/slide), and a nonsignificant trend was found toward increased urothelium (defined as >200 cells/profile) in the Cellient cell blocks (14 of 22 [64%]) compared with the histologic biopsies (17 of 44 [39%]; P = 0.070). CONCLUSIONS: Cytologic processing of ureteral microbiopsies showed superior sensitivity for detecting high-grade urothelial carcinoma, apparently owing to the increased cellular recovery.

Change in prostate cancer grade over time in men followed expectantly for stage T1c disease.

PURPOSE: We assessed whether the Gleason grade changes in men followed expectantly with nonpalpable prostate cancer diagnosed on needle biopsy (stage T1c). MATERIALS AND METHODS: We studied 241 men with stage T1c prostate cancer who were treated expectantly with repeat yearly needle biopsy sampling to assess for cancer progression. Following the initial cancer diagnosis all men had at least 1 other biopsy demonstrating cancer. RESULTS: Median patient age was 66 years. The number of biopsies showing cancer over time was 2 in 119 (49.4%), 3 in 74 (30.7%), 4 in 33 (13.7%) and 5 or greater in 15 (6.2%). The average followup for those without progression was 32.3 months. Of 241 cases 45 (18.7%) showed a significant change in grade from Gleason score 6 or less to Gleason score 7 or greater (Gleason score 7 in 41 cases, Gleason score 8 in 4 cases). Of 45 (53.5%) cases 24 that showed progression did so within 24 months of diagnosis. CONCLUSIONS: Within the first 3 years after diagnosis of Gleason score 6 prostate cancer, there is a relatively low risk of grade progression. Within the first 3 years, our data suggest that in most cases tumor grade did not evolve but rather that the higher grade component was not initially sampled since most grade changes occurred relatively soon after biopsy. Grade progression does appear to occur in some men with long-term followup who had multiple biopsies showing Gleason score 6 followed by higher grade cancer.

Characterizing sarcoma dominance pattern in uterine carcinosarcoma: Homologous versus heterologous element.

OBJECTIVE: To examine significance of sarcoma dominance (SD) patterns in uterine carcinosarcoma (UCS). METHODS: This is a secondary analysis of multicenter retrospective study examining women with stages I-IV UCS who underwent primary surgery. SD was defined as >50% of sarcoma component in uterine tumor. SD patterns were grouped as homologous sarcoma without SD (homo/non-dominance, n = 351), heterologous sarcoma without SD (hetero/non-dominance, n = 174), homologous sarcoma with SD (homo/dominance, n = 175), and heterologous sarcoma with SD (hetero/dominance, n = 189), and correlated to tumor characteristics and survival. RESULTS: SD patterns were significantly associated with age, body habitus, carcinoma type, tumor size, depth of myometrial invasion, and nodal metastasis (all, P < 0.05). On univariate analysis, SD was associated with decreased progression-free survival (PFS) and cause-specific survival (CSS) in homologous cases (both, P < 0.05) but not in heterologous cases. On multivariate models, both homologous and heterologous SD patterns remained independent prognostic factors for decreased PFS (adjusted-hazard ratio [HR] ranges: homo/dominance 1.35-1.69, and hetero/dominance 1.47-1.64) and CSS (adjusted-HR ranges: 1.52-1.84 and 1.66-1.81, respectively) compared to homo/non-dominance (all, P < 0.05). Among stage I-III disease, when tumors had SD, adding radiotherapy to chemotherapy was significantly associated with improved PFS (adjusted-HR: homo/dominance 0.49, and hetero/dominance 0.45) and CSS (0.36 and 0.31, respectively) compared to chemotherapy alone (all, P < 0.05); contrary, this association was not observed with absence of SD (all, P > 0.05). CONCLUSION: In UCS, SD impacts survival in homologous but not in heterologous type. Regardless of sarcoma types, SD was associated with decreased survival in UCS; adding radiotherapy to chemotherapy may be an effective postoperative strategy.

The role of P501S and PSA in the diagnosis of metastatic adenocarcinoma of the prostate.

BACKGROUND: Adenocarcinoma of the prostate can present as metastatic carcinoma with no known primary. Prostatic origin can be confirmed in most of these cases by immunohistochemistry for prostate-specific antigen (PSA) and prostate-specific acid phosphatase. In a small subset of high-grade prostate carcinomas, both markers are negative and therefore are not helpful for confirming prostatic origin. Recently, novel marker proteins that are preferentially expressed in prostate tissue were identified. One such marker is P501S or prostein, a 553-amino acid protein that is localized to the Golgi complex. It is expressed in both benign and neoplastic prostate tissues, but not in any other normal or malignant tissue examined to date. Owing to its apparent specificity, prostein may be a good marker to demonstrate prostatic origin in metastatic prostate cancer. DESIGN: Five-micron sections of a tissue microarray were subjected to immunohistochemistry with a monoclonal mouse anti-P501S (clone 10E3, Dako, Carpintera, CA) antibody and a monoclonal mouse anti-PSA (clone ER-PR8, Dako, Carpintera, CA) antibody. The tissue microarray contains 78 cases of metastatic prostatic adenocarcinoma, 20 cases of primary prostatic adenocarcinoma, and 20 cases of benign prostate tissue from the peripheral zone as well as samples of benign brain, pancreas, kidney, thyroid, testis, skeletal muscle, and fibroconnective tissue. RESULTS: Similar staining (intensity and extent) was identified for both markers in the majority of metastatic tumors (11 distant sites, 42 pelvic lymph nodes), in all 20 primary tumors and in all benign prostate and nonprostate tissues. The P501S stain had perinuclear cytoplasmic (Golgi) distribution even in poorly differentiated tumors and metastases. Two distant metastases were negative for PSA but retained focal weak positivity for P501S. Two other distant metastases were weakly PSA positive, but strongly P501S positive. Metastases in the pelvic lymph nodes were positive for both markers in 53 cases and 1 lymph node metastasis was strongly PSA positive but P501S negative. In summary, 67 of the 69 cases (97%) of metastatic prostate carcinomas were PSA positive, whereas 68 of the 69 cases showed at least focal weak reactivity for P501S (99%). None of the tumors were negative for both markers. CONCLUSIONS: Immunohistochemistry for P501S is a sensitive and highly specific marker for identifying prostate tissue. The large majority of metastatic prostatic adenocarcinomas are P501S positive (99%). A small subset of metastatic prostatic adenocarcinoma shows significant differences in staining intensity and extent for PSA and P501S and, therefore, combined use of these markers may result in increased sensitivity for detecting prostatic origin.

Poorly differentiated colorectal carcinoma with invasion restricted to lamina propria (intramucosal carcinoma): a follow-up study of 15 cases.

Invasive colorectal carcinomas (CRCs) with invasion confined to the lamina propria (LP) [intramucosal carcinoma (IMC)] lack access to lymphatics and therefore have no potential for metastases and local intervention (usually polypectomy) should be adequate treatment. For this reason, they are classified as "Tis" in the TNM system. It is believed that carcinomas invading the submucosa with unfavorable histology (tumors at/near the margin, and/or vascular invasion, and/or poor differentiation) require additional intervention after polypectomy, whereas those with favorable histology can be safely treated endoscopically. However, there are few data on poorly differentiated (PD) carcinomas showing invasion confined to the LP. Polypectomy is theoretically curative but in practice this has not been well demonstrated. Thus, the clinicopathologic features of 15 cases of PD CRCs with invasion limited to the LP on initial biopsies were studied to determine the best course of management for this rare subset of carcinomas. A computer search and histologic review of cases seen at Johns Hopkins Hospital was performed. Fifteen cases of PD CRC with invasion limited to the LP were identified. The clinicopathologic features of these tumors were reviewed. All 15 cases showed PD IMC with single cells infiltrating only the LP. Patients were 38 to 79 years (median, 62) of age with a male predominance (M:F=4:1). Three cases had signet ring cell differentiation, 1 had focal small cell features, and another had focal squamous differentiation. Fourteen of the cases were associated with background adenomas or adenomalike lesions including: 7 involving tubulovillous or villous adenomas, 6 involving tubular adenomas, 1 involving dysplasia associated with chronic inflammatory bowel disease. Nine of the lesions had surrounding high-grade dysplasia. One case showed no background dysplasia or adenoma. One patient was lost to follow-up and the remaining 14 were followed for 1 to 96 months (mean, 21.3 mo; median, 13 mo). Seven patients had no residual disease on follow-up colonoscopy, and no resection was performed. The remaining 7 patients were treated with partial colectomy (6) or low anterior resection (1), and of these, 5 had no infiltrating carcinoma and negative lymph nodes. One patient had a separate large colorectal (T3) carcinoma with 8/10 positive regional lymph nodes; the IMC seen on biopsy was presumably a metastasis as it was unassociated with an in situ component. Finally, the resected rectum from which an IMC had been previously detected had no residual invasive carcinoma, but the anal skin was involved by Paget disease. Thus, of the 15 cases of PD CRCs limited to the LP, 1 was a metastasis from a separate CRC and another had associated Paget disease of the anal skin. As such, even in the setting of PD carcinomas, no metastatic disease was seen arising from any of the cases that were confirmed as early primary lesions. These preliminary findings suggest that patients with isolated intramucosal PD CRCs may be managed endoscopically.

Do sporadic Peutz-Jeghers polyps exist? Experience of a large teaching hospital.

Most types of sporadic gastrointestinal (GI) polyps vastly outnumber their syndromic counterparts. In contrast, the incidence of sporadic Peutz-Jeghers polyps (PJP) is unknown. We examined all potential PJP seen at our hospital over a 22-year (y) period to assess the incidence of sporadic PJP. The pathology database of a large hospital was searched for "Peutz-Jeghers polyp(s)," yielding 121 polyps from 38 patients. The polyps were reviewed by 3 pathologists to confirm the diagnosis. Clinical information to confirm or refute a diagnosis of Peutz-Jeghers syndrome (PJS) was collected. Of the 102 polyps included after histologic review, 94 polyps arose in patients meeting the World Health Organization criteria for PJS. These PJS polyps were eliminated from further analysis. Clinical information was obtained for the remaining 8 patients with potential "sporadic" PJP (1 to 50 y; mean=14 y; median=4 y). Of the 8 potential sporadic PJP, only 3 polyps from 3 patients had unequivocal PJP histologic features, all from the small intestine. All 3 patients had clinical histories suggesting syndromic PJP although they did not meet World Health Organization criteria, that is, 2 developed pancreatic cancer, 1 had bilateral "ovarian cystic masses" and a glomus tympanicum tumor, and 1 had strong family history of GI malignancies. The 5 remaining patients each had a colonic polyp with features suggestive, but not definitely diagnostic of, PJP. In these cases, prolapse lesions could not be excluded. One patient had a history of high-grade dysplasia in a tubulovillous adenoma in the colon at 53 years, but no family cancer history. Another had a family GI cancer history. Another had a history of pituitary adenoma at age 39, and the last had ductal breast carcinoma diagnosed 4 years before the discovery of the polyp. Our findings suggest that if sporadic PJP exist, they are extremely rare. Moreover, our data suggest that individuals with a single PJP may have a cumulative lifetime risk of cancer similar to those with the syndrome.

Sessile serrated adenomas with low- and high-grade dysplasia and early carcinomas: an immunohistochemical study of serrated lesions "caught in the act".

Sessile serrated adenomas (SSAs) show serrations typical of hyperplastic polyps but display architectural differences and lack traditional dysplasia. SSAs with foci of low- (LGD) or high-grade dysplasia (HGD) or early invasive carcinoma are seldom biopsied and, thus, are not well studied. Immunohistochemical analysis for MLH1, MSH2, MSH6, and PMS2 (mismatch repair gene products) was performed on colon biopsy specimens from 11 patients (age range, 54-87 years; 4 men and 7 women) showing SSA with LGD (n = 1), HGD (n = 5), or focal invasive carcinoma (n = 5). All 11 cases showed intact nuclear staining for MSH2 and MSH6 in the SSA component; in foci of LGD, HGD, or carcinoma; and in background normal mucosa. In contrast, there was tandem loss of MLH1 and PMS2 in zones of LGD (1/1) or HGD (3/5) and early carcinoma (2/4; with concordant loss in associated HGD) but retention in SSA areas (11/11) and normal mucosa (11/11). No patient was known to have hereditary nonpolyposis colorectal cancer/Lynch syndrome. This study offers additional strong evidence that SSA is truly a precursor to at least a subset of sporadic microsatellite-unstable colorectal cancer.

The c-myc x E2F-1/p21 interactive gene expression index augments cytomorphologic diagnosis of lung cancer in fine-needle aspirate specimens.

Morphological analysis of cytologic samples obtained by fine-needle aspirate (FNA) or bronchoscopy is an important method for diagnosing bronchogenic carcinoma. However, this approach has only about 65 to 80% diagnostic sensitivity. Based on previous studies, the c-myc x E2F-1/p21WAF1/CIP1 (p21 hereafter) gene expression index is highly sensitive and specific for distinguishing normal from malignant bronchial epithelial tissues. In an effort to improve sensitivity of diagnosing lung cancer in cytologic specimens, we used Standardized Reverse Transcriptase Polymerase Chain Reaction (StaRT-PCR) to measure the c-myc x E2F-1/p21 index in cDNA samples from 14 normal lung samples (6 normal lung parenchyma and 8 normal bronchial epithelial cell [NBEC] biopsies), and 16 FNA biopsies from 14 suspected tumors. Based on cytomorphologic criteria, 11 of the 14 suspected tumors were diagnosed as bronchogenic carcinoma and three specimens were non-diagnostic. Subsequent biopsy samples confirmed that the three non-diagnostic samples were derived from lung carcinomas. The index value for each bronchogenic carcinoma was above a cut-off value of 7000 and the index value of all but one normal sample was below 7000. Thus the c-myc x E2F-1/p21 index may augment cytomorphologic diagnosis of bronchogenic carcinoma biopsy samples, particularly those considered non-diagnostic by cytomorphologic criteria.