Using Deep Learning to Predict Final HER2 Status in Invasive Breast Cancers That are Equivocal (2+) by Immunohistochemistry.

Invasive breast carcinomas are routinely tested for HER2 using immunohistochemistry (IHC), with reflex in situ hybridization (ISH) for those scored as equivocal (2+). ISH testing is expensive, time-consuming, and not universally available. In this study, we trained a deep learning algorithm to directly predict HER2 gene amplification status from HER2 2+ IHC slides. Data included 115 consecutive cases of invasive breast carcinoma scored as 2+ by IHC that had follow-up HER2 ISH testing. An external validation data set was created from 36 HER2 IHC slides prepared at an outside institution. All internal IHC slides were digitized and divided into training (80%), and test (20%) sets with 5-fold cross-validation. Small patches (256×256 pixels) were randomly extracted and used to train convolutional neural networks with EfficientNet B0 architecture using a transfer learning approach. Predictions for slides in the test set were made on individual patches, and these predictions were aggregated to generate an overall prediction for each slide. This resulted in a receiver operating characteristic area under the curve of 0.83 with an overall accuracy of 79% (sensitivity=0.70, specificity=0.82). Analysis of external validation slides resulted in a receiver operating characteristic area under the curve of 0.79 with an overall accuracy of 81% (sensitivity=0.50, specificity=0.82). Although the sensitivity and specificity are not high enough to negate the need for reflexive ISH testing entirely, this approach may be useful for triaging cases more likely to be HER2 positive and initiating treatment planning in centers where HER2 ISH testing is not readily available.

The biochemistry and regulation of S100A10: a multifunctional plasminogen receptor involved in oncogenesis.

The plasminogen receptors mediate the production and localization to the cell surface of the broad spectrum proteinase, plasmin. S100A10 is a key regulator of cellular plasmin production and may account for as much as 50% of cellular plasmin generation. In parallel to plasminogen, the plasminogen-binding site on S100A10 is highly conserved from mammals to fish. S100A10 is constitutively expressed in many cells and is also induced by many diverse factors and physiological stimuli including dexamethasone, epidermal growth factor, transforming growth factor-α, interferon-γ, nerve growth factor, keratinocyte growth factor, retinoic acid, and thrombin. Therefore, S100A10 is utilized by cells to regulate plasmin proteolytic activity in response to a wide diversity of physiological stimuli. The expression of the oncogenes, PML-RARα and KRas, also stimulates the levels of S100A10, suggesting a role for S100A10 in pathophysiological processes such as in the oncogenic-mediated increases in plasmin production. The S100A10-null mouse model system has established the critical role that S100A10 plays as a regulator of fibrinolysis and oncogenesis. S100A10 plays two major roles in oncogenesis, first as a regulator of cancer cell invasion and metastasis and secondly as a regulator of the recruitment of tumor-associated cells, such as macrophages, to the tumor site.

Plasminogen receptor S100A10 is essential for the migration of tumor-promoting macrophages into tumor sites.

Macrophages are critical drivers of tumor growth, invasion, and metastasis. Movement of macrophages into tumors requires the activity of cell surface proteases such as plasmin. In this study, we offer genetic evidence that plasminogen receptor S100A10 is essential for recruitment of macrophages to the tumor site. Growth of murine Lewis lung carcinomas or T241 fibrosarcomas was dramatically reduced in S100A10-deficient mice compared with wild-type mice. The tumor growth deficit corresponded with a decrease in macrophage density that could be rescued by intraperitoneal injection of wild-type but not S100A10-deficient macrophages. Notably, macrophages of either genotype could rescue tumor growth if they were injected into the tumor itself, establishing that S100A10 was required specifically for the migratory capability needed for tumor homing. Conversely, selective depletion of macrophages from wild-type mice phenocopied the tumor growth deficit seen in S100A10-deficient mice. Together, our findings show that S100A10 is essential and sufficient for macrophage migration to tumor sites, and they define a novel rate-limiting step in tumor progression.

The role of the annexin A2 heterotetramer in vascular fibrinolysis.

The vascular endothelial cells line the inner surface of blood vessels and function to maintain blood fluidity by producing the protease plasmin that removes blood clots from the vasculature, a process called fibrinolysis. Plasminogen receptors play a central role in the regulation of plasmin activity. The protein complex annexin A2 heterotetramer (AIIt) is an important plasminogen receptor at the surface of the endothelial cell. AIIt is composed of 2 molecules of annexin A2 (ANXA2) bound together by a dimer of the protein S100A10. Recent work performed by our laboratory allowed us to clarify the specific roles played by ANXA2 and S100A10 subunits within the AIIt complex, which has been the subject of debate for many years. The ANXA2 subunit of AIIt functions to stabilize and anchor S100A10 to the plasma membrane, whereas the S100A10 subunit initiates the fibrinolytic cascade by colocalizing with the urokinase type plasminogen activator and receptor complex and also providing a common binding site for both tissue-type plasminogen activator and plasminogen via its C-terminal lysine residue. The AIIt mediated colocalization of the plasminogen activators with plasminogen results in the rapid and localized generation of plasmin to the endothelial cell surface, thereby regulating fibrinolysis.

Regulation of fibrinolysis by S100A10 in vivo.

Endothelial cells form the inner lining of vascular networks and maintain blood fluidity by inhibiting blood coagulation and promoting blood clot dissolution (fibrinolysis). Plasmin, the primary fibrinolytic enzyme, is generated by the cleavage of the plasma protein, plasminogen, by its activator, tissue plasminogen activator. This reaction is regulated by plasminogen receptors at the surface of the vascular endothelial cells. Previous studies have identified the plasminogen receptor protein S100A10 as a key regulator of plasmin generation by cancer cells and macrophages. Here we examine the role of S100A10 and its annexin A2 binding partner in endothelial cell function using a homozygous S100A10-null mouse. Compared with wild-type mice, S100A10-null mice displayed increased deposition of fibrin in the vasculature and reduced clearance of batroxobin-induced vascular thrombi, suggesting a role for S100A10 in fibrinolysis in vivo. Compared with wild-type cells, endothelial cells from S100A10-null mice demonstrated a 40% reduction in plasminogen binding and plasmin generation in vitro. Furthermore, S100A10-deficient endothelial cells demonstrated impaired neovascularization of Matrigel plugs in vivo, suggesting a role for S100A10 in angiogenesis. These results establish an important role for S100A10 in the regulation of fibrinolysis and angiogenesis in vivo, suggesting S100A10 plays a critical role in endothelial cell function.

S100A10 regulates plasminogen-dependent macrophage invasion.

The plasminogen activation system plays an integral role in the migration of macrophages in response to an inflammatory stimulus, and the binding of plasminogen to its cell-surface receptor initiates this process. Although previous studies from our laboratory have shown the importance of the plasminogen receptor S100A10 in cancer cell plasmin production, the potential role of this protein in macrophage migration has not been investigated. Using thioglycollate to induce a peritoneal inflammatory response, we demonstrate, for the first time, that compared with wild-type (WT) mice, macrophage migration across the peritoneal membrane into the peritoneal cavity in S100A10-deficient (S100A10(-/-)) mice was decreased by up to 53% at 24, 48, and 72 hours. Furthermore, the number of S100A10-deficient macrophages that infiltrated Matrigel plugs was reduced by 8-fold compared with their WT counterpart in vivo. Compared with WT macrophages, macrophages from S100A10(-/-) mice demonstrated a 50% reduction in plasmin-dependent invasion across a Matrigel barrier and a 45% reduction in plasmin generation in vitro. This loss in plasmin-dependent invasion was in part the result of a decreased generation of plasmin and a decreased activation of pro-MMP-9 by S100A10-deficient macrophages. This study establishes a direct involvement of S100A10 in macrophage recruitment in response to inflammatory stimuli.