Inflammatory, regulatory, and autophagy co-expression modules and hub genes underlie the peripheral immune response to human intracerebral hemorrhage

Abstract

Background: Intracerebral hemorrhage (ICH) has a high morbidity and mortality. The peripheral immune system and cross-talk between peripheral blood and brain have been implicated in the ICH immune response. Thus, we delineated the gene networks associated with human ICH in the peripheral blood transcriptome. We also compared the differentially expressed genes in blood following ICH to a prior human study of perihematomal brain tissue.

Methods: We performed peripheral blood whole-transcriptome analysis of ICH and matched vascular risk factor control subjects (n = 66). Gene co-expression network analysis identified groups of co-expressed genes (modules) associated with ICH and their most interconnected genes (hubs). Mixed-effects regression identified differentially expressed genes in ICH compared to controls.

Results: Of seven ICH-associated modules, six were enriched with cell-specific genes: one neutrophil module, one neutrophil plus monocyte module, one T cell module, one Natural Killer cell module, and two erythroblast modules. The neutrophil/monocyte modules were enriched in inflammatory/immune pathways; the T cell module in T cell receptor signaling genes; and the Natural Killer cell module in genes regulating alternative splicing, epigenetic, and post-translational modifications. One erythroblast module was enriched in autophagy pathways implicated in experimental ICH, and NRF2 signaling implicated in hematoma clearance. Many hub genes or module members, such as IARS, mTOR, S1PR1, LCK, FYN, SKAP1, ITK, AMBRA1, NLRC4, IL6R, IL17RA, GAB2, MXD1, PIK3CD, NUMB, MAPK14, DDX24, EVL, TDP1, ATG3, WDFY3, GSK3B, STAT3, STX3, CSF3R, PIP4K2A, ANXA3, DGAT2, LRP10, FLOT2, ANK1, CR1, SLC4A1, and DYSF, have been implicated in neuroinflammation, cell death, transcriptional regulation, and some as experimental ICH therapeutic targets. Gene-level analysis revealed 1225 genes (FDR p < 0.05, fold-change > |1.2|) have altered expression in ICH in peripheral blood. There was significant overlap of the 1225 genes with dysregulated genes in human perihematomal brain tissue (p = 7 × 10^-3). Overlapping genes were enriched for neutrophil-specific genes (p = 6.4 × 10^-08) involved in interleukin, neuroinflammation, apoptosis, and PPAR signaling.

Conclusions: This study delineates key processes underlying ICH pathophysiology, complements experimental ICH findings, and the hub genes significantly expand the list of novel ICH therapeutic targets. The overlap between blood and brain gene responses underscores the importance of examining blood-brain interactions in human ICH.

Keywords: Autophagy; Co-expression networks; Gene expression; Gene networks; Hematoma; Hematoma clearance; ICH; Intracerebral hemorrhage; NRF2; Src kinase inhibitors.

Fig. 1

A dendrogram of the genes and their module assignments (a); a dendrogram of only the module eigengenes (representative of the overall gene expression within that module) (b). The significant modules in respect to Dx (cyan, darkolivegreen, greenyellow, magenta, red, sienna3, and tan) are marked with asterisks in (a) and colored in module-corresponding colors in (b). Dx-specific modules with significant overlap/enrichment with cell-specific genes are labeled in (a) with the specific cell type, and the hypergeometric probability p value of overlap between the module’s gene list and the cell type-specific gene list from Watkins et al. [31] and T cell-specific genes from Additional file 2: Table S1 from Chtanova et al. [32] is presented

Fig. 2

Greenyellow module. a VisANT Network of greenyellow (neutrophil- and monocyte- specific) module. The hub genes are colored in green-yellow. Neutrophil-specific hub genes are listed below. b Overrepresented canonical pathways in greenyellow (IPA analysis, Benjamini-Hochberg (B-H) corrected p < 0.05). X-axis: negative Log₁₀-transformed p-value for significance for enrichment in specific IPA pathways. Pathways higher than the black vertical line (-Log₁₀(B-H p))  > 1.3 (which corresponds to p(B-H p) < 0.05) are significant. c, d Tan module. Hub genes are colored in tan

Fig. 3

a, b VisANT network of magenta (T cell specific) module. The Hub genes are colored in magenta. T cell specific hub genes are listed below. b Overrepresented canonical pathways in magenta (IPA analysis, B-H corrected p < 0.05). X-axis: negative Log₁₀-transformed p value for significance for enrichment in specific IPA pathways. Pathways higher than the black vertical line (-Log₁₀(B-H p))  > 1.3 (which corresponds to p(B-H p) < 0.05) are significant. c, d Red module. Hub genes are colored in red

Fig. 4

a, b VisANT network of cyan (erythroblast-specific) module. The hub genes are colored in cyan. b Overrepresented canonical pathways in cyan (IPA analysis, B-H corrected p < 0.05). X-axis: negative Log₁₀-transformed p value for significance for enrichment in specific IPA pathways. Pathways higher than the black vertical line (-Log₁₀(B-H p))  > 1.3 (which corresponds to p(B-H p) < 0.05) are significant. c VisANT network of sienna3 module. Hub genes are colored in sienna3. No IPA pathways passed B-H corrected p < 0.05. d VisANT network of darkolivegreen module. Hub genes were colored in dark olive green. No IPA pathways passed B-H corrected p < 0.05

Fig. 5

LCK network in the magenta module. Genes colored in magenta are hub genes. Note other Src kinases (like FYN and ITK) and Src kinase-associated protein (SKAP1) are circled

Fig. 6

SNRNP200 network in the red module. Genes colored in red are hub genes. Yellow highlighted genes are transcription regulators

Fig. 7

STAT3 network in the tan module. Genes colored in tan are hub genes

Fig. 8

Module significant to diagnosis (Dx) and relevant gene ontology (GO) biological processes. Top 3 relevant biological processes presented in each bar. X-axis: negative Log₁₀-transformed p value for significance with Dx for each significant module. Linear p(Dx)-value displayed in each bar. Significant over-representation for cell-type specific genes displayed above each module. *GO process, which passes nominal p < 0.05, but not FDR p < 0.05; the rest pass FDR p < 0.05

Fig. 9

Overlap between differentially expressed genes in peripheral blood (current study) and perihematomal brain tissue [15] between patients with ICH and controls. There was significant enrichment (p = 6.4 × 10⁻⁰⁸) of neutrophil-specific genes in the overlap of 46 genes. The underlined genes are neutrophil-specific (based on the HeamAtlas [24]). * SATB1 (SATB Homeobox 1)—T cell-specific based on [25]. ** UBE2J1 (ubiquitin conjugating enzyme E2 J1)—B lymphocyte-specific based on [24]. *** UPP1 (uridine phosphorylase 1)—Monocyte-specific based on [24]. The green yellow genes ALOX5 (arachidonate 5-lipoxygenase) and PYGL (glycogen phosphorylase L) are hub genes in the greenyellow ICH module. The red gene AAK1 (AP2 Associated Kinase) is a hub gene in the red ICH module