Faculty Bibliography

Advancements in clinical treatment are increasingly constrained by the limitations of supervised learning techniques, which depend heavily on large volumes of annotated data. The annotation process is not only costly but also demands substantial time from clinical specialists. Addressing this issue, we introduce the S4MI (Self-Supervision and Semi-Supervision for Medical Imaging) pipeline, a novel approach that leverages advancements in self-supervised and semi-supervised learning. These techniques engage in auxiliary tasks that do not require labeling, thus simplifying the scaling of machine supervision compared to fully-supervised methods. Our study benchmarks these techniques on three distinct medical imaging datasets to evaluate their effectiveness in classification and segmentation tasks. Notably, we observed that self-supervised learning significantly surpassed the performance of supervised methods in the classification of all evaluated datasets. Remarkably, the semi-supervised approach demonstrated superior outcomes in segmentation, outperforming fully-supervised methods while using 50% fewer labels across all datasets. In line with our commitment to contributing to the scientific community, we have made the S4MI code openly accessible, allowing for broader application and further development of these methods. The code can be accessed at https://github.com/pranavsinghps1/S4MI .

Background/Purpose: The Accelerating Medicines Project (AMP) has enabled significant increases in understanding of SLE nephritis pathology, providing a profile of dozens of leukocyte subsets within affected kidneys by single-cell RNA sequencing of nephritis biopsies. While these results suggest a complex network of interactions between cell populations during nephritis, the spatial positioning of these cells is lost during the sequencing process. Inferred interactions between the diverse identified cell types would be greatly strengthened by detailed spatial information, placing these cells in context with each other and with the surrounding structures of the kidney.
Method(s): In consultation with AMP, we have used CODEX, a multicycle imaging technology allowing for staining of up to 40 targets on a single tissue sample without tissue degradation, to capture preliminary images of the AMP tissue biopsies available at New York University. Extensive antibody screening, sample preparation, optimization of antigen retrieval, and imaging steps are required, which remain under active optimization to allow for imaging the entirety of the AMP biopsy cohort available at the Grossman School of Medicine, which has been fully processed for future staining.
Result(s): At present we are able to image sixteen targets capturing dense interstitial T and B cell infiltrates, intratubular and interstitial myeloid populations, and sparser glomerular infiltrating cells in our demonstration cohort, with clear imaging of the glomeruli, tubules, and interstitial spaces. Further targets will be added as they are optimized, further allowing subsetting of T, B, and myeloid populations, with the goal of capturing the populations identified previously in single-cell sequencing.
Conclusion(s): CODEX imaging of renal biopsy samples provides spatial context for prior observations across a range of SLE nephritis samples, with complex interstitial populations found around glomeruli and tubules in active disease. Deeper profiling with expanded antigen targets to enable further sub-population phenotyping and activation states and imaging of the full cohort of biopsies available at NYU will provide a spatial atlas to SLE nephritis and further reveal underlying mechanisms of disease

The Ikaros gene (Ikzf1) encodes a family of zinc-finger transcription factors implicated in hematopoietic cell differentiation. Here we show that Ikaros suppresses the development of basophils, which are proinflammatory cells of the myeloid lineage. In the absence of extrinsic basophil-inducing signals, Ikaros(-/-) (Ik(-/-)) mice exhibit increases in basophil numbers in blood and bone marrow and in their direct precursors in bone marrow and the spleen, as well as decreased numbers of intestinal mast cells. In vitro culture of Ik(-/-) bone marrow under mast cell differentiation conditions also results in predominance of basophils. Basophil expansion is associated with an increase in basophil progenitors, increased expression of Cebpa and decreased expression of mast cell-specifying genes Hes1 and microphthalmia-associated transcription factor (Mitf). Ikaros directly associates with regulatory sites within Cebpa and Hes1 and regulates the acquisition of permissive H3K4 tri-methylation marks at the Cebpa locus and reduces H3K4 tri-methylation at the Hes1 promoter. Ikaros blockade in cultured cells or transfer of Ik(-/-) bone marrow into irradiated Ik(+/+) recipients also results in increased basophils confirming a cell-intrinsic effect of Ikaros on basophil development. We conclude that Ikaros is a suppressor of basophil differentiation under steady-state conditions and that it acts by regulating permissive chromatin modifications of Cebpa.

Histamine is a bioactive amine that exerts immunomodulatory functions, including many allergic symptoms. It is preformed and stored in mast cells and basophils but recent evidence suggests that other cell types produce histamine in an inducible fashion. During infection, it has been suggested that neutrophils may produce histamine. We also observed that histamine is released in a neutrophil-mediated LPS-induced model of acute lung injury. Therefore, we sought to examine whether innate signals promote histamine production by neutrophils. Bone marrow-derived neutrophils stimulated with a range of TLR agonists secreted histamine in response to LPS or R837, suggesting TLR4 or TLR7 are important. LPS-driven histamine was enhanced by coculture with GM-CSF and led to a transient release of histamine that peaked at 8h post stimulation. This was dependent upon de novo synthesis of histamine, since cells derived from histidine decarboxylase (HDC) deficient mice were unable to produce histamine but did generate reactive oxygen species upon stimulation. Using pharmacological inhibitors, we show that histamine production requires PI3 kinase, which has been shown to regulate other neutrophil functions, including activation and selective granule release. However, unlike mast cells, HDC deficiency did not alter the granule structure of neutrophils, suggesting that histamine does not participate in granule integrity in these cells. Consequently, our findings establish that neutrophils generate histamine in response to a select panel of innate immune triggers and that this might contribute to acute lung injury responses.

Histamine and the histamine receptors are important regulators of a plethora of biological processes, including immediate hypersensitivity reactions and acid secretion in the stomach. In these roles, antihistamines have found widespread therapeutic applications, while the last receptor to be discovered, the H4 histamine receptor, has become a major target of novel therapeutics. Recent studies involving human genetic variance and the development of mice lacking specific receptors or the ability to generate histamine have shown roles for the histamine pathway that extend well beyond the established roles. These include identification of previously unappreciated mechanisms through which histamine regulates inflammation in allergy, as well as roles in autoimmunity, infection, and pain. As a result, antihistamines may have wider applications in the future than previously predicted.

OBJECTIVE:To assess the ability of rheumatoid arthritis (RA) fibroblast-like synoviocytes (FLS) to function as antigen-presenting cells (APCs) for arthritogenic autoantigens found within inflamed joint tissues. METHODS:Human class II major histocompatibility complex (MHC)-typed FLS were used as APCs for murine class II MHC-restricted CD4 T cell hybridomas. Interferon-gamma (IFNgamma)-treated, antigen-loaded FLS were cocultured with T cell hybridomas specific for immunodominant portions of human cartilage gp-39 (HC gp-39) or human type II collagen (CII). T cell hybridoma activation was measured by enzyme-linked immunosorbent assay of culture supernatants for interleukin-2. Both synthetic peptide and synovial fluid (SF) were used as sources of antigen. APC function in cocultures was inhibited by using blocking antibodies to human class II MHC, CD54, or CD58, or to murine CD4, CD11a, or CD2. RESULTS:Human FLS could present peptides from the autoantigens HC gp-39 and human CII to antigen-specific MHC-restricted T cell hybridomas. This response required pretreatment of FLS with IFNgamma, showed MHC restriction, and was dependent on human class II MHC and murine CD4 for effective antigen presentation. Furthermore, FLS were able to extract and present antigens found within human SF to both the HC gp-39 and human CII T cell hybridomas in an IFNgamma-dependent and MHC-restricted manner. CONCLUSION/CONCLUSIONS:RA FLS can function as APCs and are able to present peptides derived from autoantigens found within joint tissues to activated T cells in vitro. In the context of inflamed synovial tissues, FLS may be an important and hitherto overlooked subset of APCs that could contribute to autoreactive immune responses.