Cooperating, congenital neutropenia–associated Csf3r and Runx1 mutations activate pro-inflammatory signaling and inhibit myeloid differentiation of mouse HSPCs

Male d715 Csf3r mice on the C57BL/6J background have been described previously [9]. Mice were housed under pathogen-free conditions in the animal facility of Tübingen University.

Mouse bone marrow cells were isolated by flushing the long bones with ice-cold PBS. Bone marrow mononuclear cells were isolated by Ficoll–Hypaque gradient centrifugation (Amersham Biosciences) and positively selected bone marrow lin⁻ cells by immunomagnetic labeling with corresponding MACS beads (Miltenyi Biotec). Cells were counted, and viability was assessed by Trypan blue dye exclusion.

To generate RUNX1 mutants, we performed site-directed mutagenesis. As a template, we used Lego-iG/Puro-RUNX1-wt-CTAP plasmid expressing human wild type (WT) RUNX1 generously provided by Dr. Boris Fehse and Dr. Carol Stocking. Specific primers to introduce mutations p.Arg139Gly (R139G) and p.Arg174Leu (R174L) in wild type RUNX1 nucleotide sequence were designed using the QuickChange Primer Design tool (https://​www.​agilent.​com/​store/​primerDesignProg​ram.​jsp). Primer sequences are available upon request. Lego-iG/Puro-RUNX1-R139G-CTAP and Lego-iG/Puro-RUNX1-R174L-CTAP plasmids were generated using the QuickChange II Site-Directed Mutagenesis Kit (Agilent Technologies, Inc.) according to the manufacturer’s instruction. RUNX1 cDNAs (WT and two mutants) were subsequently re-cloned into lentiviral pRRL.PPT.CBX3.SFFV.hRUNX1.i2.EBFP.puro.pre vector. Transgene expression was controlled by the spleen focus–forming virus promoter (SFFV) juxtaposed to the minimal ubiquitous chromatin opening element (CBX3). EBFP translation was initiated by an internal ribosomal entry side (i2).

Lin⁻ cells were cultured in a hematopoietic stem cell expansion medium consisting of Stemline II medium supplemented with Pen/Strep, 10% FCS (Sigma-Aldrich), 1 μm dexamethasone, 100 ng/ml of mSCF, 4 ng/ml of mIL-3, 10 ng/ml of hIL-6, 40 ng/ml of murine IGF-1, and 20 ng/ml of human Flt-3L at 2 × 10⁵ cells/ml for 2 days. Cells were transduced at MOIs of 5–10 in the presence of polybrene (5 μg/ml) and re-transduced after 12–24 h. Percentage of EBFP⁺ cells was assessed by FACS.

A total of 1 × 10⁶ transduced bone marrow lin⁻ cells from d715 Csf3r mice were lysed in 200 μl 3 × Laemmli buffer, and protein was denatured for 10 min at 95 °C. Five microliter of cell lysate in Laemmli buffer was loaded per lane. Proteins were separated on a 12.5% polyacrylamide gel and transferred on a nitrocellulose membrane (GE Healthcare) (1 h, 100 V, 4 °C). The membrane was blocked for 1 h in 5% milk and incubated with primary rabbit anti-human and anti-mouse RUNX1-specific (Cell Signaling Technology #4334) or GAPDH (Cell Signaling Technology, #2118) antibody overnight (at 4 °C). After that, membranes were washed and incubated with secondary HRP-conjugated antibody (Cell signaling, #7074) for 1 h at room temperature. Pierce ECL solution (Thermo Fisher) and Amersham Hyperfilms were used to detect chemiluminescence signal of proteins.

A total of 2 × 10⁵ of transduced cells/ml were incubated for 7 days in RPMI 1640 GlutaMAX supplemented with 10% FCS, 10 ng/ml hIL-6, 5 ng/ml IL-3, 5 ng/ml GM-CSF, and 10 ng/ml G-CSF. The medium was exchanged every second day. On day 7, the medium was changed to RPMI 1640 GlutaMAX supplemented with 10% FBS, 1% penicillin/streptomycin, and 10 ng/ml G-CSF. The medium was exchanged every second day until day 11. On day 11, cells were analyzed by FACS using the following antibody: rat anti-mouse Gr-1 (BD 553128) and rat anti-mouse CD11b (BD 553312) on FACSCanto II.

A total of 1000 transduced EBFP⁺ cells were plated directly after transduction in 1 ml methylcellulose medium (MethoCult GF M3434; StemCell Technologies) supplemented with 10 ng/ml of G-CSF. After 14 days of culture, the numbers of CFU-G, CFU-GM, and BFU-E colonies were counted. Cells were collected for the colony replating experiments, washed 3 times with PBS, and plated 1000 cells/dish in new methylcellulose for an additional 2 weeks (1st replating). The procedure was repeated one more time (2nd replating).

After 48 h of lentiviral transduction, lin⁻ cells were starved for 24 h and subsequently treated with 10 ng/ml of G-CSF for 24 h. After that, transduced cells were sorted, and mRNA was isolated using the RNeasy Mini Kit (Qiagen, #74106) according to the manufacturer’s instructions.

RNA from transduced lin⁻ cells was subjected to microarray analysis using the Affymetrix Microarray Platform. The GeneChip WT cDNA Synthesis and Amplification Kit was used to make double-stranded cDNA from total RNA, which was then labeled with biotin (GeneChip WT Terminal Labeling Kit). After chemical fragmentation of the biotin-labeled cDNA targets, they were hybridized to the GeneChip Mouse Gene 2.0 ST Array using the Fluidics Station 450 and scanned using the Affymetrix GeneChip Scanner 3000 with the GeneChip Operating Software 1.4 (Affymetrix, Santa Clara, CA). Data analysis was performed using Affymetrix Expression Console Version 1.1 for invariant set normalization, and the Ingenuity Pathway Analysis (IPA) software (Qiagen) was used for identification of differentially expressed genes. Motif activity response analysis (MARA) was conducted using the Integrated System for Motif Activity Response Analysis (ISMARA); CEL files were uploaded to the server using the web interface.

After sorting, 1 × 10⁴ cells were centrifuge onto microscope slides at 250 rpm for 3 min. The slides were air-dried, and subsequent Wright–Giemsa staining was carried out. Images where acquired on a Nikon Eclipse TS100. Cells were covered with oil, and images were collected at × 630 magnification.

RNA isolation was performed using the RNeasy Micro Kit (Qiagen). cDNA was synthesized from 1 μg total RNA with the Omniscript RT Kit (Qiagen). qRT-PCR was conducted with SYBR Green qPCR master mix (Roche) on a Light Cycler 480 (Roche). Target genes were normalized to ACTB. Primers are available upon request.

A total of 2 × 10⁵ cells were incubated with FcR blocking antibody (BioLegend #101320), 7AAD, and lineage cocktail biotin–conjugated antibody (BioLegend #79750, #79748, #79752, #79748, #79749), and stained with APC-Cy7-conjugated streptavidin (BioLegend #405208), anti-mouse Sca-1 BV510–conjugated antibody (BioLegend, #108129), and anti-mouse-c-KIT APC–conjugated antibody (BioLegend, #105812). After staining and washing, cells were analyzed on a CANTO II (BD) flow cytometer.

We isolated bone marrow lin⁻ cells from d715 Csf3r mice and transduced these cells with lentiviral vectors encoding EBFP and RUNX1 wild type (RUNX1-WT), or mutants, RUNX1-R139G, or RUNX1-R174L. Amino acids 174 and 139 were found to be “hot spots” for mutation in the RHD domain of the RUNX1 protein, as they were detected in four and three CN-AML patients, respectively (Fig. 1a) [10] (data not shown). At 72 h post-transduction, we performed colony-forming units and replating experiments of sorted EBFP⁺ cells (Fig. 1b). We observed comparable expression of human WT and mutant RUNX1 proteins in transduced cells (Fig. 1c). Low levels of endogenous murine runx1 protein were detected in BFP⁺ control transduced cells (Fig. 1c). Interestingly, d715 Csf3r lin⁻ cells transduced with mutated RUNX1 had markedly diminished capacities to form myeloid colonies, including CFU-G and CFU-GM, as compared with cells transduced with WT RUNX1 (Fig. 1d). In line with these findings, CFU replating experiments showed that cell transduced with each RUNX1 mutant had markedly higher replating capacities than RUNX1 WT–transduced cells (Fig. 1e).

We next compared the G-CSF-triggered myeloid differentiation of transduced d715 Csf3r lin⁻ cells in vitro. Transduced cells were cultured in liquid culture myeloid differentiation medium for 11 days (Fig. 2a). FACS was used to count the absolute numbers of myeloid CD11b⁺ and Gr-1⁺ cells on day 11 of liquid culture, and the results showed that d715 Csf3r lin⁻ cells transduced with each of the RUNX1 mutants exhibited reduced myeloid differentiation, compared with WT RUNX1–overexpressing samples (Fig. 2b). At the same time, no marked difference was observed in the absolute numbers of BFP⁺ cells between groups transduced with WT RUNX1 or with each of RUNX1 mutants on days 3 and day 7 of liquid culture differentiation. On day 11 of culture, numbers of BFP⁺ cells transduced with RUNX1-R139G mutant were significantly (p < 0.05) increased, as compared with RUNX1 WT– or RUNX1-R174L–transduced samples (Fig. 2c).

We further aimed to identify intracellular signaling pathways downstream of csf3r and RUNX1 mutations that may affect the in vitro proliferation and myeloid differentiation of mouse HSPCs. The d715 Csf3r lin⁻ cells were expanded in HSPC expansion medium and transduced with lentiviral vectors carrying RUNX1 WT, RUNX1-R139G, or RUNX1-R174L. Transduced cells were starved for 24 h and then treated with G-CSF for 24 h. EBFP⁺ cells were sorted in RLT buffer, and mRNA expression was evaluated using an Affymetrix MoGene 2.0 Chip (Fig. 3a). Representative images of Wright–Giemsa–stained cytospins prepared from sorted cells show no difference in the cell morphology between studied groups: all samples show immature cell morphology (Fig. 3b). These data suggest that the differences in mRNA expression should not contribute to a strong diversity in the cell composition between studied groups.

After invariant set normalization, the expression levels of the RUNX1 mutants were normalized to those obtained from WT RUNX1–transduced cells. The fold change expression table (Suppl. Table 1) was uploaded to the IPA software, and the cutoff values were set to − 1.7 and + 1.7 for log fold change, with the goal of identifying differentially expressed genes. This analysis showed that 1113 and 1814 genes were differentially expressed in the RUNX1-R174L and RUNX1-R139G mutant groups compared with the WT RUNX1 group (Suppl. Table 2). Overlap analysis revealed that 679 genes (37.4% for the RUNX1-R139G mutant and 61% for the RUNX1-R174L mutant) were co-regulated in both RUNX1 mutant groups (Fig. 3c). The 15 most highly up- and downregulated genes were very similar between the RUNX1 mutant groups. Of the top upregulated genes, 10/15 from the RUNX1-R174L group and 14/15 from the RUNX1-R139G group were co-activated in both groups. Of the top downregulated genes, 14/15 from the RUNX1-R174L group and 11/15 from the RUNX1-R139G group were co-downregulated (Fig. 3d). A comparison analysis performed using the IPA software revealed that 46 and 79 canonical pathways were significantly enriched (log(p value) ≥ 1.3 corresponding to p ≤ 0.05) in d715 Csf3r lin⁻ cells transduced with RUNX1-R174L or RUNX1-R139G, respectively. Of these pathways, 11 in the RUNX1-R174L group and 18 in the RUNX1-R139G group were significantly activated (z-value ≥ 2) or inhibited (z-value ≤ − 2). Most of the regulated pathways belonged to members of the activated innate immune pathways category; these included NF-kappaB signaling, toll-like receptor signaling (TLR), acute-phase response signaling, production of nitric oxide and reactive oxygen species, pattern-recognizing receptors, Trem 1 signaling, and IL-1 signaling. Activation was also seen among members of the inflammatory signaling pathways, such as IL-6 signaling, interferon signaling, Tec kinase signaling, and leukocyte extravasation signaling (Fig. 3e, Suppl. Table 3).

As examples of the activated canonical inflammatory and innate immune signaling pathways, canonical TLR signaling and IL-6 pathways are depicted in Suppl. Figs. 2 and 3. TLR signaling was upregulated in d715 Csf3r lin⁻ cells transduced with the R139G or R174L RUNX1 mutants; this resulted in activation of transcription factors through Janus kinase 1 (JAK1) and p38 MAPK, promoting the expression of IL-1, TNF-alpha, and IL-12 (Suppl. Fig. 2A, B). Interferon signaling acted through JAK1 and JAK2 to activate the transcription factors, STAT1 and STAT2, thereby upregulating interferon response factors and other pro-inflammatory genes (Suppl. Fig. 2A, B).

The upregulation of IL-6 signaling–dependent genes in d715 Csf3r lin⁻ cells transduced with each RUNX1 mutant was also mediated through JAK signaling, but relied on the transcription factor, STAT3, to confer translational regulation in the nucleus. Additional IL-6-dependent activation of ERK1 induced the NF-kappaB-NF-IL-6 transcription factor complex (Suppl. Fig. 3A, B).

Upstream analysis performed using the IPA software identified potential upstream regulators responsible for the gene expression signatures observed in the studied groups. We detected 768 upstream regulators in the RUNX1-R139G mutant group and 875 regulators in the RUNX1-R174L mutant group (Suppl. Fig. 4A, Suppl. Table 4). Five hundred upstream regulators were commonly found in both groups: 65.1% in the RUNX1-R139G mutant group and 57.1% in the RUNX1-R174L mutant group (Suppl. Fig. 4A). Among the 15 top activated upstream regulators, 9/15 for RUNX1-R174L and 11/15 for RUNX1-R139G were shared between the groups. Among the 15 top downregulated regulators, overlaps were seen in 6/15 for RUNX1-R174L and 9/15 for RUNX1-R139G (Suppl. Fig. 4B). Selected up- and downregulated candidate genes were validated by qRT-PCR in the independent set of experiments (Suppl. Fig. 4C). Also, staining of transduced G-CSF-treated lin⁻ cells with Sca-1 and c-kit antibody followed by flow cytometry analysis revealed an increase of lin⁻Sca-1⁺c-KIT⁺ cell population in RUNX1-R139G– and RUNX1-R174L–transduced groups, compared with cells expressing RUNX1 WT, or GFP CTRL (Suppl. Fig. 4D).

The top overlapping upstream regulators included a number of inflammatory cytokines, such as IL-15 and IL-18. IL-15 has been described as a responsible driver for chronic inflammation in autoimmune diseases and hematological malignancies [20]. Both IL-15 and IL-18 are confirmed targets for antitumor activity [21, 22]. Another factor that promotes survival of leukemogenic cells, Bcl2a1 [23], was also highly expressed, as were Ly6a and Sca-1, which are known as stemness factors for HSPCs.²⁴ The top downregulated genes included anti-inflammatory genes, such as Il-10 [25], and the cytokine-processing factor, Dpp4, which can increase proliferation by prolonging cytokine signaling [26]. Interestingly, Csf3r was among the highest expressed genes in both groups. This is in accordance with the literature showing that G-CSF sensitivity is increased in AML blasts with missense RUNX1 mutations.

We used the ISMARA [27] web tool to analyze the transcription factor binding motifs enriched among the differentially expressed genes. We found that the most highly enriched motif was Irf2_Irf1_Irf8_Irf9_Irf7 (z-value 4.564), which corresponds to the interferon regulatory factors. This indicates that an interferon-related change in inflammation signaling is responsible for at least some of the observed gene expression differences. The Stat2 motif was significantly correlated with the expression signature (z-value 2.707), and we observed significant activation of the Klf4_Sp3 (z-value 2.349) and Mecp-2 (z-value 2.627) transcription factor binding motifs known to be associated with AML or MDS, but significant downregulation of the Max–Mycn motif (z-value 2.016), upon transduction of d715 Csf3r lin⁻ cells with RUNX1 mutants (Suppl. Fig. 5A). Finally, our inferred activity analysis of each motif revealed that there was a high degree of similarity in the significantly activated motifs associated with the two different RUNX1 mutants (Suppl. Fig. 5B).