Programmed cell death 1 (PD-1) and its ligand (PD-L1) are key suppressors of the cytotoxic immune response. PD-L1 expression on tumor cells may be induced by the immune microenvironment, resulting in immune escape (adaptive immune resistance), and an adverse prognosis in many malignancies. In colorectal carcinoma the response to PD-1/PD-L1 inhibition is correlated with microsatellite instability. However, little is known about the clinicopathologic, molecular, and prognostic characteristics of colorectal carcinoma with PD-L1 expression. We performed immunohistochemistry for PD-L1 on 181 cases of colorectal carcinoma with known microsatellite instability and mutational status, and correlated PD-L1 expression with clinicopathologic features including tumor-infiltrating lymphocyte burden/immunophenotype, tumor mutational profile, and disease-specific survival. PD-L1 was expressed in tumors from 16 patients (9%) who were more often older (P = 0.006) and female (P = 0.035), with tumors exhibiting a larger size (P = 0.013), but lower stage (Po0.001). PD-L1 expression was associated with increased CD8 and TBET-positive tumor-infiltrating lymphocytes, medullary phenotype, poor differentiation, microsatellite instability, BRAF mutation (Po 0.001 for each), and a lower frequency of KRAS mutation (P = 0.012). On multivariate analysis, PD-L1 expression was associated with medullary morphology and frequent CD8-positive tumor-infiltrating lymphocytes, suggesting adaptive immune resistance. PD-L1 positivity was not predictive of survival in the entire cohort, but it was associated with a lower disease-specific survival within the microsatellite-instability high cohort. PD-L1 expression in colorectal carcinoma is associated with clinicopathologic and molecular features of the serrated pathway of colorectal carcinogenesis, and is associated with a worse outcome within microsatellite-unstable tumors. These findings support the role of PD-L1 expression in providing normally immunogenic colorectal carcinoma a means of immune evasion and a more aggressive biology, provide a potential mechanistic explanation for the favorable response of microsatellite-unstable colorectal carcinoma to PD-1/PD-L1 pathway blockade, and suggest potential predictive and prognostic roles of PD-L1 immunohistochemistry in colorectal carcinoma.

Immunoglobulin G4-related disease (IgG4-RD) is an uncommon disorder that demonstrates characteristic clinicopathologic features including sclerosing lesions with storiform fibrosis, increased IgG4+ plasma cells with an increased IgG4+/IgG+ plasma cell ratio, obliterative phlebitis, and often an increased serum IgG4 level. This review summarizes the characteristic histopathologic and clinical features of IgG4-RD with detailed discussion of the histopathologic characteristics of the most commonly involved anatomic sites. We also present recent advances in our understanding of the pathophysiologic mechanisms of IgG4-RD and discuss updates on the treatment, prognosis, and outcomes of this rare disease, including discussion of the possible association between IgG4-RD and malignancy.

In colorectal carcinoma the evaluation of BRAF mutation status is increasingly being performed given its utility as a prognostic and predictive biomarker. However, there are conflicting reports of the sensitivity and specificity of BRAF V600E immunohistochemistry, and little is known about its reliability in tissues collected from metastatic sites or following chemo/radiation or targeted therapy. The degree of intratumoral staining heterogeneity is also not well established. We performed immunohistochemistry for BRAF V600E (VE1) on 204 cases of colorectal carcinoma including 59 with the BRAF V600E mutation. These included primary (n=147) and metastatic/recurrent (n=57) tumors, collected before (n=133) or after (n=71) chemo/radiation or targeted therapy. Evaluation of a test cohort (39 cases) with knowledge of mutation status established a specific staining pattern for the mutation: diffuse cytoplasmic staining of near-uniform intensity, regardless of strength of staining. Using this pattern, pathologists at three levels of training independently performed blinded evaluation of the remaining cases. BRAF V600E staining was 96.3% sensitive and 98.5% specific for the mutation, including both pre- and post-treatment specimens. Fleiss’ kappa for interobserver agreement was 0.96. Staining of whole sections of the BRAF mutants showed diffuse staining in all cases and uniform or near-uniform intensity in 91%. In 20 cases with both pre- and post-treatment specimens there was 100% accuracy and agreement in staining between samples. We conclude that BRAF V600E immunohistochemistry is reliable for the evaluation of mutational status in colorectal carcinoma regardless of site or prior treatment history, and staining shows a high degree of intratumoral homogeneity.

Diagnostic histopathology is a gold standard for diagnosing hematopoietic malignancies. Pathologic diagnosis requires labor-intensive reading of a large number of tissue slides with high diagnostic accuracy equal or close to 100 percent to guide treatment options, but this requirement is difficult to meet. Although artificial intelligence (AI) helps to reduce the labor of reading pathologic slides, diagnostic accuracy has not reached a clinically usable level. Establishment of an AI model often demands big datasets and an ability to handle large variations in sample preparation and image collection. Here, we establish a highly accurate deep learning platform, consisting of multiple convolutional neural networks, to classify pathologic images by using smaller datasets. We analyze human diffuse large B-cell lymphoma (DLBCL) and non-DLBCL pathologic images from three hospitals separately using AI models, and obtain a diagnostic rate of close to 100 percent (100% for hospital A, 99.71% for hospital B and 100% for hospital C). The technical variability introduced by slide preparation and image collection reduces AI model performance in cross-hospital tests, but the 100% diagnostic accuracy is maintained after its elimination. It is now clinically practical to utilize deep learning models for diagnosis of DLBCL and ultimately other human hematopoietic malignancies.