Cluster of differentiation 19 (CD19)-directed immunotherapies such as chimeric antigen receptor (CAR) T-cells and bispecific T-cell engagers (BITEs) are known to be clinically effective in treating B-cell malignancies, but they are also known to cause a high incidence of neurotoxicity. Therefore, exploring the etiologies of CAR T-cell toxicities affecting patient outcomes is an important area of investigation.
In a recently published study in Cell, Parker et al.1 used multimethod analyses on several datasets including single-cell ribonucleic acid sequencing (scRNA-seq) of human brain cells to potentially identify causes of neurotoxicity following immunotherapy using CD19-targeted CARs.
Methods
- scRNA-seq analysis across multiple human datasets and whole mouse brain
- Samples included 2,364 cells from the human prefrontal cortex, 7,906 from the human forebrain, and 1,977 from the human ventral forebrain
- Immunohistochemistry (IHC) analysis of several regions of the brain using samples from healthy deceased subjects
- Transcriptome analysis of pericytes and vascular smooth muscles cells (VSMCs) using a scRNA-seq dataset generated by the BRAIN Initiative Cell Census Network (BICCN), including approximately 857,000 cells across 101 individual samples
- Bulk RNA-seq data from different human age-groups and brain regions generated by the Allen Institute’s BrainSpan project, containing prenatal (n = 237) and postnatal (n = 287) samples from diverse brain regions
- Genome-wide correlation analysis across postnatal samples
Results
scRNA-seq
- Sequencing data obtained from 2,364 human prefrontal cortex cells using various markers (PDGFRB, FOXF2, RGS5, CD248, CDH5, and PECAM1) allowed segregation into distinct populations such as astrocytes, lymphocytes, microglia cells, oligodendrocyte precursors, endothelial cells, and pericytes
- Within these cell clusters, a small population of cells representing around 1.5% of all nonneuronal coexpressed CD19 and the mural cell marker CD248, were identified; these cells did not express the VSMC marker ACTA2 nor the B-cell-specific marker CD79A
- The level of CD19 expression in pericytes from the human prefrontal cortex ranked in the 86th and 71st percentiles and was similar to other pericyte marker genes such as CD248, RGS5, and PDGFRB (85th, 96th, and 98th percentiles, respectively).
IHC
- Expression of CD19 protein was confirmed by IHC of several regions of the human brain, in cells located in the abluminal areas close to vasculature consistent with mural cells. Abluminal CD19 staining was not found along all vessels, suggestive of interpatient differences
- Although generally rare, CD19-positive cells were found to be more abundant across several brain regions, specifically, in the hippocampus, insula, temporal lobe, and parietal lobe
- CD19 expression along smaller, as well as larger, blood vessels suggests additional expression in VSMCs
Transcriptome analysis of pericytes
- Transcriptome analysis showed that CD19 was highly expressed in neurovascular meta-cell clusters lacking expression of B-cell markers, indicative of pericytes and VSMCs
- Further analysis of a CD19-expressing nonneuronal subset (mural cells, endothelial cells, and microglia) demonstrated strong enrichment of pericyte markers, ABCC9 and KCNJ8
- Analysis using the BICCN dataset identified early progenitors biased towards nonneuronal lineages. Gene expression analysis confirmed the staining pattern observed by IHC, suggesting that CD19 expression may be frequent in human brain mural cells
Bulk RNA-seq
- Bulk RNA-seq analysis showed that CD19 was expressed in both prenatal and postnatal samples at similar levels and expressed in samples from different brain regions
- Genome-wide analysis revealed correlation of CD19 gene expression with CD248, CSPG4, ANPEP, FOXS1, and FN1 gene expression which reconfirmed that CD19 in the brain results from the presence of mural cells rather than B cells
Analysis using mouse models
- Analysis of mouse brain cells by flow cytometry suggested the presence of both a CD45-high CD19+ B-cell population and a rarer CD45−CD19+ cells population
- scRNA-seq of cells isolated from separated whole mouse brain confirmed the low level CD19 expression in a population of CD13+CD45− cells identified as pericytes
- However, these CD19+ pericytes were relatively less abundant in the mouse brain compared with the human brain
- To answer the question whether targeting CD19 by CAR T-cell infusion in mice would lead to neurological changes, the authors infused CD19-directed CAR T-cells into immunodeficient, nontumor-bearing mice. 7 days after infusion, the mice that had received mouse CD19-targeting CAR T-cells showed increased blood-brain barrier (BBB) permeability as measured by Evans Blue dye, while the mice that had received the human CD19-targeting CAR T-cells did not
- Furthermore, only mice that had been infused with the mouse CD19-targeting CAR T-cells demonstrated decreased numbers of both CD45+CD19+ B cells and CD45−CD13+CD19+ mural cells
Mural specific expression of CD19
- Comparative analysis of human brain pericytes with pericytes and VSMCs from the lung, revealed numerous transcriptional differences between brain and lung pericytes
- CD19 was notably expressed in the brain, but not in the lung mural cells
Key findings
- scRNA-seq revealed CD19 expression in human brain mural cells
- Mural cells line blood vessels and maintain the BBB integrity
- Neurotoxicity using CD19-directed CAR T-cell therapy may be mediated by targeting CD19+ mural cells
Conclusion
As CAR T-cell therapy is at the therapeutic forefront in treating leukemia and lymphoma, it is crucial to understand the risks and benefits of these treatments for patients. This study adds to the understanding of the biology of CAR T-cells in patients and provides a strong stimulus for further investigating the potential targeting of mural cells by CD19-directed CAR T-cell therapies, in order to develop therapies with improved safety profiles.
The findings of this study are aligned with previously established principles on the response to CAR T-cell therapy and neurotoxicity. However, the study does have several limitations, as it did not determine the precise contributions of mural cell death and cytokine release syndrome in this process, and it did not investigate whether there were interpatient differences in mural cell frequency and/or CD19 expression. Also, neurovascular cell heterogeneity and long-term effects associated with the targeting of mural cells was not analyzed.
The study highlights the need for developing a comprehensive human single-cell gene expression atlas for clinical medicine. Rare cell types might be easily missed in measurements of bulk tissue due to their low frequency, but as this study demonstrates, they could be critically important in understanding the clinical effects of targeted therapies.