Transcripts of repetitive DNA elements signal to block phagocytosis of hematopoietic stem cells

Abstract

Macrophages maintain hematopoietic stem cell (HSC) quality by assessing cell surface Calreticulin (Calr), an "eat-me" signal induced by reactive oxygen species (ROS). Using zebrafish genetics, we identified Beta-2-microglobulin (B2m) as a crucial "don't eat-me" signal on blood stem cells. A chemical screen revealed inducers of surface Calr that promoted HSC proliferation without triggering ROS or macrophage clearance. Whole-genome CRISPR-Cas9 screening showed that Toll-like receptor 3 (Tlr3) signaling regulated b2m expression. Targeting b2m or tlr3 reduced the HSC clonality. Elevated B2m levels correlated with high expression of repetitive element (RE) transcripts. Overall, our data suggest that RE-associated double-stranded RNA could interact with TLR3 to stimulate surface expression of B2m on hematopoietic stem and progenitor cells. These findings suggest that the balance of Calr and B2m regulates macrophage-HSC interactions and defines hematopoietic clonality.

Fig.1.. Chemical screen identifies ROS^dependent and ROS^Independent surface Calr-inducers.

(A) Representative imaging from 3 independent experiments showing that ROS+Runx1+ cells are preferentially doomed by the macrophages. Green: Mpeg; Red: Runx1; white: CellRox probe. (B) ROS levels in embryonic HSPCs marked by surface Calreticulin (Calr). Data were analyzed by Pearson correlation. MFI, median fluorescence intensity. A.U, arbitrary unit. Data points represent a pool of 100–300 3 dpf embryos (C) Surface Calr levels on HSPCs following ROS inhibition with diphenylene (DPI) or VAS2870 (Nox inhibitor). Data were analyzed by Kruskall-Wallis test followed by Dunn’s; *P <0.05. DMSO, dimethyl sulfoxide Data points represent a pool of 100–300 3 dpf embryos. Data are means ± SEM (A-C) were performed in a pool of 100–300 zebrafish embryos in 2 independent experiments. (D) A schematic overview of the chemical screen: Two approaches were employed to ensure that our system specifically recorded the surface Calreticulin values 1) HEK293 cells were transfected with the SPLIT-TURBO ID where the C-Terminus was design to carry Cdh2, a membrane protein, while the N-terminus was carrying Calr, and 2) Anti-Calr was conjugated with a A647 using the Zenon technology. Next, the cells were plated in 384-well plates and treated with a panel of 1200-bioactive small molecules in 3 different concentrations (5uM, 2.5uM, and 0.625 uM). The cells were also labeled with a ubiquitous ROS probe (CellRox) and DAPI for nuclear staining. After 24 hrs the Zeiss Cell Discoverer 7 (CR7) microscope was used to read the plates. Data were analyzed using the CellProfiler and R-Sight HTS software. SD, Split-turbo ID. ab, Calreticulin antibody. DMSO, dimethyl sulfoxide. (E) Numbers and overlap in compounds that induced surface Calreticulin in HEK293 for both systems. Each compound was tested in replicates and 2–4 random fields of view were chosen for acquisition. The same well was used for the SPLIT-Turbo, Ab-Calr and CellROX acquisition. (G) Left panel: Number of compounds that increased surface Calreticulin in vitro and in vivo (zebrafish embryos). Right panel: Live image microscopy was used to determine the Interaction fraction between macrophage-HSPCs using mpeg-GFP;runx1-mCherry embryos that label macrophage and HSPCs, respectively. Highlighted in blue are the ROS^indepedent and black represent ROS^dependent compounds. Data are means ± SEM. Calr, surface Calreticulin. ROS, Reactive oxygen species. The zebrafish live imaging experiments performed to determine the macrophage-HSPCs interaction were conducted as two independent experiments and a minimum of 5 embryos were imaged.

Fig.2. Tlr3 is required for HSPC grooming.

(A)Stack Plot showing the macrophage interaction behavior in the presence of the Calr-inducers. rGrooming behavior is shown in the black bar, while dooming behavior as the orange bar. The imaging was conducted using 3 dpf zebrafish embryos. The exact sample size is indicated in the figure, n=6–16 from 2 independent experiments. Data were analyzed by Wilcoxon matched-pairs signed rank test. *P=0.0156, **P=0.002, ***P=0.001. (B) A scheme overview of the Gecko 2.0 CRISPR/Cas9 knockout screen. (C) sgRNAs significantly enriched (P<0.05) were identified using the MAGECK software and plotted as cumulative stack based on their Z-Score. The arrows indicate examples of significant enriched sgRNAs and their Z-Score depicted inside the brackets. The screen was conducted using K562 cells. The screen was performed in 3 independent experiments. (D) Knock-down of Tlr3 (Tlr3 MO) or iTlr3 treatment reduces the fraction of HSPCs that are groomed, but increases the fraction of HSPCs that are doomed in 3 dpf zebrafish embryos. The macrophage behavior percentage outcome was calculated by counting the total number of interacting macrophages and dividing by the total number of dooming or grooming events. Data points represent an image embryo, SD MO n=9, TL3 MO n= 11, DMSO n=4 and iTLR3 n=5. Experiments were performed in 2 independent experiments. Two-way ANOVA followed by Sidak’s multiple comparison. **P=0.0027; ***P=0.00005. (E) Representative imaging showing the experimental design of the parabiosis experiment. (F) Both macrophages showed higher interaction rates with the HSPCs. *P=0.04, **P=0.0015. Data were analyzed using a two-way ANOVA test. The macrophage-HSPCs interaction percentage was calculated by counting the total number of HSCPs (runx1+ cells) and dividing by the total number HSCPs interacting with a macrophage. (G) Bar-plot showing the macrophage behavior outcome fraction. The macrophage behavior percentage outcome was calculated by counting the total number of interacting macrophages and dividing by the total number of dooming or grooming events. *P= 0.01, **P=0.0057. Data were analyzed using an unpaired Student t-test. (F-G). Data points represent an image embryo, n= 10 from 3 independent experiments.

Fig.3. B2m acts as a “don’t eat-me” molecule and prevails unwarranted HSPCs dooming.

(A) b2m knock-out stable mutants (B2m KO) showed significant reduction in B2M levels. Data were analyzed by unpaired Mann Whitney test. *P=0.028. Data points representing a pool of 100 embryos (B)Barplot showing the interaction percentage. b2m KO embryos showed significant lower macrophage-HSPCs interaction with a (C-D) dooming dominance. Data were analyzed by unpaired Mann-Whitney. *P=0.012, **P=0.0021, ****P<0.0001. Data points represent an image embryo. Wildtype n=12 and B2m KO n=14. This data represents 3 independent experiments. The macrophage-HSPCs interaction percentage was calculated by counting the total number of HSCPs (runx1+ cells) and dividing by the total number HSCPs interacting with a macrophage. (E) Representative imaging showing the increased dooming behavior observed in the B2m homozygous mutant. n=4 from 2 independent experiments (F) Bar-plot showing the interaction rates in wildtype and B2m KO zebrafish embryos, n=4. Data points depict the field of view. (G) The macrophage behavior percentage outcome was calculated by counting the total number of interacting macrophages and dividing by the total number of dooming or grooming events. We found that while in the wildtype background DL-PPMP promotes grooming, in the B2m KO background it promotes dooming. (H) A scheme overview of the murine HSPCs (LSK+) and macrophage co-culture. Left lower panel is a representative imaging showing the sorted murine HSPC-MHC-I⁺ (Green) being groomed by a macrophage (magenta), while on the right lower panel is a murine HSPC-MHC-I^- being doomed by a macrophage. (I) MHC-I⁺ HSPCs showed higher grooming rates compared to MHC-I^- cells. Data were analyzed using non parametric one-way ANOVA followed by Kruskall Wallis. *P=0.03. Data are means ± SEM. The macrophage behavior percentage outcome was calculated by counting the total number of interacting macrophages and dividing by the total number of dooming or grooming events. (J) A schematic overview of the Zebrabow-M system: Animals with 15 to 20 insertions of a multicolor fluorescent cassette are crossed to the draculin:CreERT2 line to enable clonal labeling of lateral plate mesoderm lineages. By treating with 4-hydroxytamoxifen (4-OHT) at 24 hpf just after HSC specification, individual stem cell lineages express specific fluorescent hues that can be quantified in the adult marrow. Families of Zebrabow-M;draculin:CreERT2 animals injected with b2m morpholino with or without clodronate liposomes exhibit (K) reduced numbers of HSC clones. (L-M) Clonal dominance in the adult marrow. Clodronate liposome was injected at 48 hpf. Data points represent an adult kidney marrow. Control n= 22, b2m sgRNA n=21 and clodronate n=3. Experiments were conducted at least 2 times. (A, B, C, D,F,G,I,K,L,M) Data are means ± SEM. Experiments were conducted in vivo using zebrafish as a model. (N) A schematic overview of the clonal diversity in regards to the “don’t eat-me” signal.

Fig.4.. Repetitive elements as Erv elicit a virus-like response in HSPCs promoting B2m expression.

(A) Flow cytometry shows that ~ 20% of runx1+23mCherry+ HSPCs expresses isg15 (gating strategy is shown in fig. S4A). Data points represent a pool of ~200 embryos acquired in 3 independent experiments. Data are means ± SEM. Statistical analysis performed by Mann-Whitney test. P = 0.003. (B) Positive correlation between B2m and Isg15 levels. Data were analyzed by Pearson correlation. Data points represent a pool of ~200 embryos acquired in 3 independent experiments. (C) Left panel representative image of a Isg15+-HSPC interaction. HSPCs, red; macrophages, blue;isg15, green. Right panel: HSPCs-Isg15+ cells showed lower dooming ratios. (D) Sorted isg15+HSPCs showed higher b2m and RE expression. Data points represent a pool of ~200 embryos. Experiments were performed 2 times. Y-axis depicts the relative gene expression (E) Uniform manifold and projection (UMAP) of sorted runx1+23-mcherry HSPCs in standard morpholino (gray, Control sdM) and irf8 depleted embryos (yellow, irf8 sdKD) (Original data from (6)). (F) Cluster 2 (C2), enriched for control HSPCs, showed higher expression of RE. We analyzed 808 cells for this analysis. (G) Treating 2 dpf embryos with CM272 (5 uM) led to higher number of isg15+HSPCs in the 3dpf CHT. Data was quantified by live cell imaging Data were analyzed by Mann-Whitney test. *P=0.02. Data points represent an embryo. Experiments were performed 2 times. (H) EdU staining of runx1+23:mCherry embryos treated with CM272 identifies a significant increase in proliferating HSPCs at 3dpf. Data were analyzed by Student’s t-test. Data normality was enquired by the Shapiro-Wilk test. **P=0.0061. Data are means ± SEM. Control n= 5 and CM272 n=7. (I) CM272 treated embryos showed lower dooming ratios. Data points represent an embryo. Percentage was calculated by quantification of Edu+ Runx1+ cells divided by the total number of runx1+ cells. The plot depicted 2 independent experiments. Control n=6 and CM272 n=10 (J) Gini coefficient in control (DMSO), CM272 and CM272 b2m knockdown zebrafish. Data were analyzed by Kruskall-Wallis test followed by Dunn’s multiple comparison test. **P=0.0043. Data points represent an adult fish. DMSO n=6, CM272 n=5 and b2m sgRNA n=5. The plot depicted 2 independent experiments.(K) b2m-TWISTR Zebrabow-M;draculin:CreERT2 treated CM272 exhibits reduced numbers of HSC clones in the adult marrow. Data were analyzed by Kruskall-Wallis test followed by Dunn’s multiple comparison test. *P=0.023. Data points represent an adult fish. DMSO n=6, CM272 n=5 and b2m sgRNA n=5. The plot depicted 2 independent experiments. (L) ISG15 expression positively correlates to the transposable elements (TE) expression in human HSCs. Previously published single cell RNA-seq datasets of human bone marrow CD34+ stem and progenitors (original data from (90), n = 4 samples) were aligned to a TE reference using CellRanger v3.2. TE expression was normalized by total RNA counts. Pearson correlation was performed between TE expression and ISG15 gene expression of individual HSCs, represented in each datapoint. (M) Human bone marrow HSCs from the same dataset as in (L) were grouped based on their ISG15 expression, where in the upper quartile was considered the threshold for high and the lower quartile as low. Normalized expression of B2M is shown for individual cells. ISG15 high cells showed higher B2M expression. Only cells with at least one transcript of ISG15 were included in the analysis. High, n = 203 cells, low n = 608 cells, from n = 4 samples; P = 0.02 using a linear mixed model, with sample as a random effects variable. Box plot represents the median, bottom and top quartiles, whiskers correspond to 1.5 × the interquartile range. (N) Human Erv overexpression upregulated the B2M levels on human CD34+ cells. Data are means ± SEM. Data points depict a CD34 donor. The data collected from 2 independent (O) EdU staining of MFG+ mice treated with DL-PPMP identifies an increase in proliferating HSPCs 24 hrs after the treatment. Percentage was determined by the total number of living cells. Data points depict a mouse. (P) DL-PPMP treatment stimulates B2M expression in murine HSPCs. Flow cytometry was performed using the cells from Tibia/Femur (gating fig. S4Q). Data were analyzed by Student’s t-test. *P=0.02. Data are means ± SEM. n mice = 3 per condition. (Q) A schematic overview of the proposed molecular pathway resulting in higher surface B2m.