Presenter
Miguel Medina-Serpas (University of Florida)

Authors
Miguel Medina-Serpas, Gregory Golden, Maigan Brusko, Trevor Rogers, Richard Musca, Michael Betts, Klaus Kaestner, Ali Naji, Mark Atkinson, Todd Brusko

Purpose
Spatial transcriptomic technologies enable in situ whole transcriptome detection thereby preserving biologically relevant signals typically lost in disaggregated single cell approaches – rare cells, stromal cell populations, and cell:cell interactions. This is advantageous in studying the histopathology of type 1 diabetes (T1D), where both immune and pancreatic cells have been implicated in initiation and progression of insulitis, and where striking islet and lobular heterogeneity has been described within the pancreas of T1D organ donors. To investigate key features of disease pathogenesis, we acquired formalin-fixed paraffin-embedded (FFPE) sections from paired pancreas and pancreatic lymph node (pLN) of 16 total donors (Mean Age: 20.25 y.o.[ (SD: +/- 6.12)(Range: 10 – 31 y.o.)], 37.5% Female) and employed the Visium Spatial Gene Expression Array (10x Genomics). Resultant spatial transcriptome data was linked with annotated single cell RNAseq data from matched tissues and individuals to bolster cell type annotation.

Methods
We analyzed a cross-sectional cohort comprised of non-diabetic controls (n = 6), at-risk subjects presenting with a single (n = 3) or multiple (n = 2) autoantibodies (AABs), and T1D (n = 6) subjects with variable disease duration (mean+SD: 4.67 yr +3.20 yr). The Visium Spatial Gene Expression Array contains defined capture areas (6.5mm x 6.5mm or 11.5mm x 11.5mm capture areas), composed of a grid of 55µm diameter capture spots containing a probe-based amplification system . Spatial gene expression libraries were pooled at equimolar ratios and sequenced at minimum to 25,000 reads/ capture spot according to manufacturer protocol. Low quality capture spots were identified and removed based on total number of reads, total number of detected genes, and <10% mitochondrial gene content. Gene counts were normalized and scaled for interdonor differential expression analysis. Visium capture spots are. To overcome this technical limitation, in silico cell deconvolution was performed using a single-cell resolution reference, and tissue functional regions (e.g. islets, follicles) were annotated.

Summary of Results
In silico cell deconvolution of pancreas spatial data revealed an immune cell signature within at-risk and T1D donors compared to non-diabetic controls, with greater enrichment detected among donors that present with multiple AABs and/or closer to disease onset. Furthermore, we identified increased inflammatory chemokine expression (CXCL12, CCL2) (LogFC ≥ 0.25), with significant enrichment of CCL2 (p = 0.0245, Wilcoxon rank sum test) in the pancreatic islets of the aforementioned subject groups. Differential expression analysis of inflammation-associated genes revealed significant increases in TNFRSF1A, TNFAIP3, and IFNGR2 expression (p ≤ 0.05, Wilcoxon rank sum test) in the pancreatic islets, with parallel increases observed in downstream signaling gene networks. Finally, preliminary analysis of paired pLN data via cell deconvolution has revealed enrichment of cell-type specific gene expression programs at specific stages of T1D. Notably, we identified relative enrichment of Tcm/Treg GEX programs in non-diabetic controls, as well as effector CD8+ T-cell programs in at-risk and T1D subjects. Differential expression analysis of pLN revealed upregulation of various chemotactic (CXCL13, CCL19, CCR7) (LogFC ≥ 0.25) and immune modulatory genes (ZAP70, TNFRSF4), and significant upregulation of TXNIP (p = 0.036, Wilcoxon rank sum test) within the T-cell zone of the pLN.

Conclusions
Collectively, we demonstrate the potential of emerging technology amd bioinformatic tools to unvocer unique insights in the immune histopathological features of T1D. Application of in silico cell deconvolution to pancreas spatial data identified a distinct immune cell signature is enriched in at-risk and T1D donors presenting with multiple islet-specific autoantibodies relative to non-diabetic controls. We further identified a global and compartment-level inflammatory chemokine signature in donor pancreata which follows a similar pattern of enrichment to the identified bulk immune signature which may represent a non-specific mechanism contributing to pancreatic immune infiltration in T1D. Lastly, we identified cell-type specific GEX programs and disease stage patterns of enrichment within matched donor pLN which we intend to leverage in resolving specific transcriptional phenotypes from the bulk pancreas signature. Ongoing work includes validating ST signature genes and pathways in serially prepared tissue sections using high parameter single cell resolution probe and protein staining technologies which we expect will provide novel leads for new therapeutic targets for therapeutic interventions in T1D.