CRISPRi screens in human astrocytes elucidate regulators of distinct inflammatory reactive states

Abstract

In response to central nervous system injury or disease, astrocytes become reactive, adopting context-dependent states with altered functions. Certain inflammatory insults induce reactive astrocyte states that lose homeostatic functions and gain harmful outputs, and likely contribute to neuroinflammatory and neurodegenerative diseases. However, the cellular pathways controlling these states are not fully understood. Here, we combined single-cell transcriptomics with CRISPRi screening in human iPSC-derived astrocytes to systematically interrogate inflammatory reactivity. We found that autocrine-paracrine IL-6 and interferon signaling downstream of canonical NF-kappaB activation drove two distinct inflammatory reactive states promoted by and inhibited by STAT3, respectively. Furthermore, these states corresponded with those observed in other experimental contexts, including in vivo, and their markers were upregulated in the human brain in Alzheimer's disease and hypoxic ischemic encephalopathy. These results and the platform we established have the potential to guide the development of therapeutics to selectively modulate different aspects of inflammatory astrocyte reactivity.

Meta-analysis of human and mouse ALS astrocytes reveals multi-omic signatures of inflammatory reactive states

“…To assess whether ALS astrocytes display shared gene expression changes with A1 astrocytes, we next compared the pan-ALS transcriptomic signature with treated A1-reactive astrocytes. We used five hiPSC-derived astrocyte protocols, each with TNF-, IL1A-, and C1q (A1)-treated and vehicle (A0)-treated samples from Barbar et al (2020) (one protocol: three A1 and three A0 lines) and Leng et al (2021) (four protocols: each with three A1 and three A0 lines) (Supplemental Fig. S4A,B).…”

“…Similarly, meta-analysis of mouse models of ALS astrocytes (SOD1 G93A , Tardbp, and Tmem259 depletion) was calculated by comparing the knockout (or mutant) versus control groups, controlling for data set variation in the design formula (∼data set + group). Results for proinflammatory factor-treated A1 astrocytes were created by comparing A1 (TNF-, IL1A-, and C1q-treated) versus A0 (untreated) samples, among five distinct protocols from Barbar et al (2020) and Leng et al (2021). The samples from the protocol named Fernandopulle were removed because this was an outlier protocol on the PCA and A1 samples clustered together with vehicle-treated samples.…”

“…Astrocytes adopt unique reactive states depending on the context-specific stimuli that they encounter, which can have detrimental or protective effects on neurons (Anderson et al 2016;Liddelow et al 2017;Escartin et al 2021). Incubating astrocytes in the proinflammatory factors, TNF, IL1A, and complement pathway component C1q (encoded by the C1qa/b/c genes and referred to as C1q hereafter), induces a deleterious reactive state, referred to as "A1" (Liddelow et al 2017), with impairment of homeostatic functions including glutamate uptake in human induced pluripotent stem cell (hiPSC)-derived and mouse astrocytes (Liddelow et al 2017;Barbar et al 2020;Guttenplan et al 2020;Leng et al 2021). Triple knockout of Tnf, Il1a, and C1qa slowed disease progression in the SOD1 G93A mouse model of ALS, directly implicating inducers of an A1-reactive state in ALS (Guttenplan et al 2020).…”

Chemical inhibition of pathological reactive astrocytes promotes neural protection

“…These pathogenic reactive astrocytes are found in neurodegenerative diseases including Alzheimer's disease, Parkinson's disease, Huntington's disease, amyotrophic lateral 2 sclerosis, vanishing white matter disease 10 , and multiple sclerosis 1,11 . Thus, there is considerable interest in modulating pathological reactive astrocytes to reduce the progression of these diseases 3,6,12 . Here, we develop an astrocyte discovery platform and leverage the power of highthroughput phenotypic drug screening 13 to identify modulators of reactive astrocytes.…”

“…Finally, libraries were sequenced by the Case Western Reserve University Genomics Core on an Illumina HiSeq2500 with paired-end 50bp reads and a target sequence depth of 50,000 reads per cell. Sequence data were first processed by 10x Cell Ranger v3.1 using default settings to generate a gene expression matrix and then all downstream analysis was performed with the R package, Seurat v4.0 12 .…”