Soluble PD-L1 reprograms blood monocytes to prevent cerebral edema and facilitate recovery after ischemic stroke

Abstract

Acute cerebral ischemia triggers a profound inflammatory response. While macrophages polarized to an M2-like phenotype clear debris and facilitate tissue repair, aberrant or prolonged macrophage activation is counterproductive to recovery. The inhibitory immune checkpoint Programmed Cell Death Protein 1 (PD-1) is upregulated on macrophage precursors (monocytes) in the blood after acute cerebrovascular injury. To investigate the therapeutic potential of PD-1 activation, we immunophenotyped circulating monocytes from patients and found that PD-1 expression was upregulated in the acute period after stroke. Murine studies using a temporary middle cerebral artery (MCA) occlusion (MCAO) model showed that intraperitoneal administration of soluble Programmed Death Ligand-1 (sPD-L1) significantly decreased brain edema and improved overall survival. Mice receiving sPD-L1 also had higher performance scores short-term, and more closely resembled sham animals on assessments of long-term functional recovery. These clinical and radiographic benefits were abrogated in global and myeloid-specific PD-1 knockout animals, confirming PD-1+ monocytes as the therapeutic target of sPD-L1. Single-cell RNA sequencing revealed that treatment skewed monocyte maturation to a non-classical Ly6C^lo, CD43^hi, PD-L1+ phenotype. These data support peripheral activation of PD-1 on inflammatory monocytes as a therapeutic strategy to treat neuroinflammation after acute ischemic stroke.

Keywords: Immune checkpoints; Inflammation; Ischemic stroke; Large vessel occlusion; MCAO; Monocytes; sPD-L1.

Fig. 1.. PD-1 was upregulated on circulating monocytes after stroke.

(A) Representative patient images of an acute right MCA occlusion on CTA, and associated MRI images demonstrating a large MCA territory infarct (ADC/DWI) with associated edema (T2). (B) Flow cytometric analysis showing frequencies of monocyte subsets in the blood during the acute period after MCA occlusion and reperfusion. (C) Changes in PD-1 expression on blood monocytes after MCA occlusion and reperfusion. (D) PD-1 mean fluorescence intensity (MFI) on blood monocytes after MCA occlusion and reperfusion in a cohort of 14 patients. (E, F) Correlation analysis of maximum and average PD-1 MFI on circulating monocytes with edema-to-infarct ratio. (G) Correlation analysis of maximum and average PD-1 MFI on monocyte subsets with edema-to infarct ratio.

Fig. 2.. PD-L1 attenuated brain injury and improved survival after MCAO.

(A) Kaplan-Meier curve. 30-day survival rate after MCAO was significantly increased by treatment with soluble PD-L1. P < 0.0022, log-rank test; n = 19 to 56 mice per experimental arm. (B, C) Bar graphs demonstrate that brain water content is increased at 48 h after MCAO. Treatment decreases water content, though not back to baseline. P < 0.0001 sham versus MCAO; p = 0.0003 sham versus MCAO + PD-L1; p = 0.0266 MCAO versus MCAO + PD-L1N = 8–10 mice per experimental group. (D). Representative MRI images of wild type mice, 72 h after MCAO. ADC and TRACE sequences show a right MCA territory infarct pattern. The T2- weighted sequence demonstrates brain edema (high signal intensity) within and surrounding the core infarct. (E) Bar graph demonstrates no significant change in infarct volume with treatment (p = 0.3295). However, treatment significantly decreases the volume of edema surrounding the core infarct (F), p = 0.0001, and volume of total edema per volume of core infarct (G), p < 0.0001. n = 14 to 15 per experimental group. Statistical analyses were performed by Mann-Whitney test. Error bars represent +/− SEM.

Fig. 3.. Soluble PD-L1 therapy improved short- and long-term functional outcomes.

(A) Bar chart shows the 28-point neuroscores at week 1 for wild-type mice categorized by quartile (1st, 2nd, 3rd, 4th). Statistical analyses were performed by Chi-square and Fisher exact tests to evaluate the association between treatment with PD-L1 and having a high 28-point neuroscore (4th quartile). P = 0.011, n = 35 to 38 per experimental group. (B) Graphical representation of select gait parameters (C, D) Bar graphs depict percent change in median when compared to sham mice. Displayed are all the variables that were significantly modified by stroke for the Digigait test at week 1 and week 3. (E) Representative tracings of the mouse’s trajectory in an open field. (C, D) Bar graphs depict percent change in median when compared to sham mice. Displayed are all the variables that were significantly modified by stroke for the Digigait test at week 1 (C) and week 3 (E), as well as the open field test at week 4 (F). Significant p < 0.05 when comparing MCAO and sham after a Bonferroni correction for multiple comparisons. (G) Table summarizes the number of variables that, with treatment, more closely resembled sham animals than untreated MCAO animals.

Fig. 4.. Brain infiltrating myeloid cells were critical mediators of the sPD-L1 treatment effect.

(A) Representative flow cytometry plots to illustrate gating strategy for PD-1 + monocytes (CD11b + CD45^hi). (B) MCAO results in significant infiltration of monocytes into the ischemic hemisphere (p = 0.0159) at 48 h after MCAO. Though the total numbers are not significantly affected by treatment, PD-L1 administration significantly decreases the frequency of PD-1 + monocytes (C) and overall expression as measured by mean fluorescence intensity (MFI) (D). Experiments run in duplicate with n ≥ 4 mice per arm. (E) Representative MRI images of global myeloid knockout (PD-1^−/−) mice and myeloid-specific knockout (PD-1^f/fLysMcre) mice, 72 h after MCAO. ADC (low signal intensity) and TRACE (high signal intensity) sequences depict the core infarct, while the T2- weighted (high signal intensity) sequence demonstrates edema. (F) Bar graph demonstrates no significant change in the volume of edema surrounding the core infarct or for the volume of total edema per volume of core infarct (G). N = 8 per group for the global PD-1 knockout arms, and n = 4 to 7 for the myeloid-specific knockout arms. Statistical analyses were performed by Mann-Whitney test. Error bars represent +/− SEM. (H-I) Bar chart depicts the 28-point neuroscores at week 1 for global and tissue-specific PD-1 knockout mice categorized by quartiles. Statistical analyses using Chi-square and Fisher exact tests demonstrated no significant association between PD-L1 treatment and having a high 28-point neuroscore (4th quartile). N = 25 to 27 per group for the global PD-1 knockout arms, and n = 4 to 8 for the myeloid-specific PD-1 knockout arms.

Fig. 5.. Soluble PD-L1 reprogrammed circulating monocytes to a restorative subtype with downregulation of inflammatory markers and increased metabolic reserve.

(A) UMAP of PD-L1 treated and untreated monocytes single cell RNA sequencing demonstrated the transcriptional heterogeneity of PD-L1 treated and untreated murine monocytes isolated at 48 h after MCAO. (B) Volcano plot showing 5,455 significantly differentially expressed genes (Benjamini-Hochberg adjusted p < 0.05). Horizontal blue line indicates pmin, which is the smallest p-value represented in R (~2.23 e-308). Points at the blue line indicate genes with p-value < pmin. (C) Top 15 positively (red) and negatively (blue) enriched gene sets from Gene Ontology: Biological Processes. (D) UMAP showing 11 transcriptional clusters derived from Louvain community detection. (E) Table demonstrating number of PD-L1 treated cells, total cells and proportion of treated cells in each cluster and p-value of differential proportion test between proportion of PD-L1 treated cells in each cluster and proportion of PD-L1 treated cells in the overall data. (F) UMAP showing expression of select genes for monocyte subset and activation markers. (G) Scatter plot showing the effects of exposing in vitro myeloid cells to anti-PD-1 antibody and/or sPD-L1 on Ly6C expression. (H) Oxygen consumption rate of myeloid cells treated with PD-L1 +/− anti-PD-1 blocking antibody (I) Bar graphs depict calculated values for respiratory parameters.