Expression analysis of Zeb1 and Zeb2 mRNA during early hematopoiesis

To determine unique as well as overlapping expression patterns for Zeb1 and Zeb2 mRNA in early murine hematopoiesis, we have reanalyzed the previously published [19] single-cell (sc)RNA-seq data with more recent bioinformatics tools [20,21]. In general, we discovered that irrespective of what mouse HSPC population is examined Zeb2 appears to be relatively higher than Zeb1 mRNA levels (Fig 1A). This is particularly true in the CMP/MEP(MMP2)/GMP(MMP3) populations. One exception to this trend is that Zeb1 mRNA levels appear to be higher in short-term HSC (ST-HSC)/multipotent progenitor (MPP1) cells prior to commitment to further myeloid and lymphoid fate specification (MPP2 and MPP4/LMPP) (Fig 1A). In addition, a larger percentage of cells within these CMP/MEP(MMP2)/GMP(MMP3) populations express Zeb2 than Zeb1 (heatmaps of sublineage frequencies in Fig 1A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Expression analysis of Zeb1 and Zeb2 during early hematopoiesis.

(A) Levels of Zeb1 (upper) and Zeb2 (lower) across LT-HSC (dark green to light green channels, Lin⁻c-Kit⁺ Sca1⁺CD34^–Flk2^–), HSPC (dark red to purple, LSK) and progenitors (dark red to light red channels, LSK), respectively. Zeb2 levels are higher throughout except for ST-HSCs where Zeb1 levels are higher (asterisk). Heatmaps at the bottom of graphs show percentage of cell lineage that express Zeb1 or Zeb2. HSPC subpopulations were subdivided into HSC1 and HSC2 (projected) expressing low/absent and high SLAM marker CD229, respectively, including ESLAM cells (CD45⁺EPCR⁺CD48⁻CD150⁺). (B) (Upper) dimensionality reduction and trajectories of 3 major cell types (LT-HSC, HSPC, and Prog) across 1,920 hematopoietic single cells sequenced by Nestorowa and colleagues [19], analyzed with the Monocle3 algorithm [20]. Clusters identified by Monocle3 were enclosed in a black box. Two major clusters consisting in LT-HSC and HSPC (see clusters 1 and 3, respectively) and 1 cluster of progenitors (see cluster 2) were detected. Also, another restricted cluster of LT-HSC were found (see cluster 4). (Lower) Pseudotime fitting analysis of these cells with tradeSeq program [21]. (C) (Left) Expression of Zeb1 gene across the single cell experiment, color coded according to the log10 of Zeb1 pseudocounts. Cells with zero pseudocounts for Zeb1 expression are colored in gray. (Right) Estimated expression obtained for Zeb1 across cell types expressing Zeb1 and arranged by pseudotime with the tradeSeq program. Across pseudotime, fitted curves indicates up-regulation of Zeb1 expression in LT-HSC (black curve), down-regulation of Zeb1 in HSPC (green curve) and mild down-regulation of Zeb1 in progenitors (yellow curve). (D) (Left) Same as left C for Zeb2 gene. (Right) Same as right C for Zeb2 gene. Across pseudotime, fitted curves indicate up-regulation of Zeb2 expression in LT-HSC and HSPC (black and green lines, respectively) and up-regulation of Zeb2 in progenitors (yellow curve). HSPC, hematopoietic stem and progenitor cell; LSK, Lin⁻Sca1⁺cKit⁺; LT-HSC, long-term HSC; ST-HSC, short-term HSC.

https://doi.org/10.1371/journal.pbio.3001394.g001

Using sc mapping and pseudotime algorithms [20,21] we can lineage trace long-term HSC (LT-HSC)➔HSPC➔Prog cell transitions with LT-HSCs (occupying the left most part of graphs) and the most differentiated GMP committed progenitors (on the far right; Fig 1B). As expected, we discovered a major cluster containing LT-HSC and HSPC in the beginning of the pseudotime that drives differentiation of more committed MMP (HSPC) and progenitors (CMP /MEP/GMP) in a single trajectory, but in parallel also gives rise to a small population of LT-HSCs. We hypothesized this cluster will serve as a self-renewal niche rather than a source for differentiation, due to its position in the terminal branches of the trajectory curve [22] (clusters 3 and 4 in Fig 1B). According to more continuous differentiation models of hematopoiesis [23], these populations readily express some myeloid as well B and T-cell markers not seen in LT-HSCs and may represent more lineage primed HSPCs.

Using these data, we could then plot Zeb1 and Zeb2 expression data versus pseudotime to understand better how Zeb1 and Zeb2 mRNA steady-state level changes are occurring throughout early hematopoietic development. Here, we could show that (1) Zeb2 mRNA levels start out high in LT-HSC and HSPC cells, then gradually fall off, but then start to steadily increase as lineage progenitors become more committed (Fig 1C); (2) Zeb1 mRNA levels increase as LT-HSCs differentiate, but then fall off quickly in HSPCs, and then gradually decrease in more committed progenitors; and (3) in addition to being more highly expressed, many more progenitors express Zeb2 than Zeb1 transcripts (Fig 1D).

From the ImmGen gene expression database (https://www.immgen.org/), there appears to be overlapping and, in some cases, complementary expression between Zeb1 and Zeb2, with Zeb1 being more highly expressed in mature B/T lymphoid cell subsets and Zeb2 more highly expressed in the monocytic/macrophage lineages (S1 Fig).

Hematopoietic-restricted genetic inactivation of Zeb1 does not lead to excessive embryonic/postnatal lethality or cephalic hemorrhage

In order to study ZEB1 in embryonic hematopoiesis, we genetically intercrossed Zeb1 conditional mice (Zeb1^fl/fl) [24] with the Tie2-Cre [25] and hematopoietic-restricted Vav-iCre [26] lines, as previously performed for Zeb2 (using Zeb12 ^fl/fl) [8]. Despite the fact that Tie2-Cre^Tg/+; Zeb1 ^fl/fl and Vav-iCre^Tg/+; Zeb1 ^fl/fl mice were present at sub-mendelian rates at weaning (7% Tie2-Cre^Tg/+; Zeb1^fl/fl versus 12.5% control Cre -ve; Zeb1^fl/fl and 20% Vav-iCre^Tg/+; Zeb1 ^fl/fl versus 25% Cre–ve; Zeb1 ^fl/fl), we observed no evidence for embryonic or postnatal lethality associated with cephalic hemorrhaging, which was observed for such Zeb2-deficient models [8] (S2A Fig): There, no viable embryos were observed at E13.5 (with Tie2-Cre), and only 3% of the mutant homozygous null pups survived at P7 (with Vav-iCre) [8] (S2B Fig). One major difference here is that the Zeb1 KO experiments were performed on a pure C57Bl/6 inbred background, whereas the original Zeb2 KO experiments [8] were originally performed on a mixed outbred background.

Hematopoietic-restricted KO of Zeb1 leads to decreased HSPC populations, decreased myeloid cell development, and altered T-cell differentiation

To document the cell-autonomous role of Zeb1 in definitive hematopoiesis, CD45.2⁺ HSPCs were isolated from E14.5 fetal livers from Tie2-Cre; Zeb1 ^fl/fl embryos and used to reconstitute lethally irradiated syngeneic C57Bl/6 CD45.1 mice in long-term reconstitution assays (Fig 2A). At 23 weeks post-reconstitution, HSPC analysis (Fig 2B) showed no changes in % or total number of SLAM marker (CD150⁺CD48⁻) LT-HSCs (Fig 2C, S3A Fig). Overall, there were decreases in % and cell number of Lin⁻Sca⁺cKit⁺ cells (Fig 2D, S3B Fig: *p < 0.05). CD135/CD34 staining of this Lin⁻Sca⁺cKit⁺ population showed decreases in % of live ST-HSCs and MPPs but no significant changes in % LT-HSCs or in overall cell numbers (Fig 2D, S3B Fig: *p < 0.05, ***p < 0.001). FCγR_II/III/CD16-CD34 analysis of the MPP population (Lin⁻cKit⁺Sca1⁻) showed significant decreases in % of CMP (CD34⁺FCγR_II/III^low) and GMPs (CD34⁺FCγR_II/III⁺), but no changes in MEPs (CD34⁻FCγR_II/III^-) as well as overall decreases in GMP numbers (Fig 2D, S3B Fig: *p < 0.05, ***p < 0.001). Moreover, decreases in CD11b⁺ myeloid cells (particularly monocytic lineage cells (i.e., CD11b⁺, LyS6G^-) were observed in the peripheral blood (PB) and bone marrow (BM) as well as decreased neutrophils (NEUs; i.e., CD11b⁺, LyS6G⁺) in the BM (Fig 2E, S3C and S3D Fig: *p < 0.05, ****p < 0.0001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Hematopoietic-specific loss of Zeb1 leads to differentiation defects in specific HSPC populations as well as myeloid lineage and T-cell defects.

(A) Schematic of BM transplant experiments using Zeb1 null CD45.2⁺ fetal liver HSPCs from Tie2-Cre, Zeb1^fl/fl mice, and Cre only controls transplanted into CD45.1⁺ recipients with PB and BM analysis conducted at 23 weeks post-transplant. (B) Overview of flow cytometry gating strategy used to analyze hematopoietic stem and progenitor (HSPC) populations. LSK cells were analyzed for SLAM marker expression (CD150, CD48) or were analyzed by parallel (CD135/CD34) marker expression to define MPP as well as ST and LT HSC populations more accurately. MPPs were analyzed by FcgammaR, CD34 expression to further define MEP, GMP, and CMP populations. (C) SLAM marker expression showing similar percentage of LT-HSCs (CD150⁺CD48⁻) in Zeb1 null and control BM. (D) Overall, there were decreases in the percentage of Lin⁻Sca⁺cKit⁺ cells likely composed of significant decreases in the percentage Zeb1-deficient BM percentages for ST-HSCs and MPPs. Moreover, there were significant decreases in the percentage of as CMP and GMP cell populations in the BM of Zeb1 null reconstituted recipients. (E) Flow cytometric analysis of PB (left panel), BM of reconstituted mice showed defects in Zeb1 null HSPC contribution to myeloid cells (Cd11b⁺) including monocytic (Cd11b⁺Ly6G) and NEU (Cd11b⁺Ly6G⁺) lineage cells. (F) Cytometric analysis of thymic T-cell populations showed significantly decreased percentage of CD25⁻CD44⁻ DN4 progenitors and increased CD8⁺ mature T cells. Error bars indicate SD of the mean (n = 4 per group, *p < 0.05, **p < 0.01, ****p < 0.0001, nonparametric t test). Raw data behind graphs are included in A in S1 Data. BM, bone marrow; HSPC, hematopoietic stem and progenitor cell; LSK, Lin⁻Sca1⁺cKit⁺; LT-HSC, long-term HSC; MPP, multipotent progenitor; NEU, neutrophil; PB, peripheral blood; ST-HSC, short-term HSC.

https://doi.org/10.1371/journal.pbio.3001394.g002

Given the previous defects observed in T-cell differentiation in Zeb1 global deletion mice [11], we also examined T-cell development in these transplant settings after long-term reconstitution. We could demonstrate significantly decreased thymocyte numbers with significant decreases mainly in % of live DN4 cells (i.e., CD25⁻, CD44⁻) as well as increased CD8⁺ skewing in the thymus (Fig 2F: **p < 0.01, ***p < 0.001). No significant changes were observed in overall B cell differentiation (B220 marker analysis; S3E Fig).

Overall, these results are slightly in contrast to hematopoietic Zeb2 null mice that showed enhanced granulocytic differentiation, as well as increased HSPC populations including MEPs, suggesting potentially divergent roles of Zeb1 and Zeb2 in controlling multilineage hematopoietic differentiation [9].

Zeb1 null HSPCs show decreased self-renewal and multilineage differentiation potential

To examine the cell-autonomous role of ZEB1 in HSC maintenance and self-renewal, we performed serial methylcellulose replating assays of Zeb1-deficient HSPCs obtained from E14.5 fetal livers of Vav-iCre; Zeb1 ^fl/fl embryos as well as Cre-negative controls. We have previously demonstrated that the Vav-iCre line is specific to the hematopoietic lineage at this stage, whereas the Tie2-Cre system is active in both the endothelium and hematopoietic system [8]. Here, we demonstrated decreased hematopoietic colony formation potential of Zeb1-deficient HSPCs; in primary plating experiments there was a 50% reduction in myeloid lineage colony numbers (Fig 3A). This decrease in colony numbers was exacerbated in secondary plating experiments whereby the Zeb1-deficient HSPCs gave rise to 80% to 90% fewer myeloid lineage colonies compared to controls (Fig 3A: *p < 0.05, ***p < 0.001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Hematopoietic-restricted loss of Zeb1 decreases hematopoietic colony formation potential and ability to compete with wt HSPCs for contribution to all hematopoietic lineages.

(A) HSPCs isolated from Vav-iCre; Zeb1^fl/fl fetal livers show decreased numbers of colonies in methylcellulose-based colony assays (first) that further decreases upon secondary replating (second) compared to Cre negative controls. (B) Schematic of competitive BM reconstitution experiments whereby equal numbers of BM cells from Zeb1 null CD45.2⁺ mice were mixed with equal numbers of control CD45.1⁺ cells and used to reconstitute lethally irradiated CD45.1⁺ mice. If Zeb1 null CD45.2+ HSPCs are not compromised, they would be expected to contribute equally as the control CD45.1+ cells in their contribution to all hematopoietic cells (equal CD45.2⁺-orange/CD45.1⁺-blue, top row) whereas if they are severely compromised then the control CD45.1⁺ cells will solely contribute to the reconstituted hematopoietic system (all CD45.1⁺- blue, lower row). (C) Zeb1 null (Vav-iCre, Zeb1^fl/fl) CD45.2⁺ donor cells (orange bars) were outcompeted by control CD45.1⁺ competitor HSPCs (blue bars) for their ability to contribute to all hematopoietic cells analyzed in the PB and BM (rightmost panels). Zeb1 ^fl/+ heterozygous (middle panels) and Cre negative (left panels) CD45.2⁺ doner cells in general contributed equally as well as the competitor CD45.1⁺ cells for their contribution to the hematopoietic system of recipient mice with the exception to the T-cell lineage. Data are represented as mean + SD from 4 biological replicates. *p < 0.05; **p < 0.01, nonparametric t test. Raw data for (A) are included in S1 Data. Raw data for (C) are included in B of S1 Table. BM, bone marrow; HSPC, hematopoietic stem and progenitor cell; PB, peripheral blood; wt, wild-type.

https://doi.org/10.1371/journal.pbio.3001394.g003

To further demonstrate that Zeb1 null HSPCs are deficient in self-renewal and multilineage differentiation potential, we performed competitive BM repopulation assays. Here, equal numbers of BM derived HSPCs from Vav-iCre^Tg/+; Zeb1 ^fl/fl homozygous, Vav-iCre^Tg/+; Zeb1^fl/+ heterozygous and Cre-negative Zeb1 ^fl/fl CD45.2⁺ mice were mixed with equal numbers of competitor HSPCs from wt CD45.1⁺ BM. These BM mixtures were used to reconstitute lethally irradiated CD45.1⁺ recipient mice (Fig 3B). CD45 cytometric analysis of the PB and BM of long-term reconstituted BM recipients clearly demonstrated that Zeb1-deficient donor HSPCs (CD45.2⁺) were outcompeted by wt competitor cells (CD45.1) in their capacity to give rise to all mature hematopoietic cells including B, T-, as well as myeloid cells in the PB and BM (Fig 3C). The overall percentages of chimerism and lineage-specific chimerism for these experiments are summarized in A–C of S1 Table.

These results are similar to the results obtained from the previous analysis of Zeb2-deficient HSPCs in competitive reconstitution experiments: These demonstrated that Zeb2 null HSPCs could be outcompeted by wt competitor cells for their contribution to all mature hematopoietic cells, with the exception of granulocytes where there was some contribution of Zeb2-deficient CD45.2⁺ progenitor cells [9].

DKO of Zeb1 and Zeb2 causes PB cytopenia and severe differentiation defects in HSPCs

Given this similarity in Zeb2- and Zeb1 null hematopoietic cells in being outcompeted by wt control HSPCs for their ability to contribute to multilineage hematopoiesis, we next examined if inducible loss of both Zeb1 and Zeb2 may exacerbate the severity of observed hematopoietic phenotypes of the individual gene KOs. For this, we used the tamoxifen-inducible Rosa26-Cre-ERT2 [27] system that was bred into conditional Zeb1 ^fl/fl; Zeb2 ^fl/fl [28] and compound Zeb1 ^fl/fl; Zeb2 ^fl/fl backgrounds (Fig 4A). To validate this inducible system, we used R26-Cre-ERT2; Zeb2 ^fl/fl and R26-Cre-ERT2; Zeb1 ^fl/fl BM cells in transplant experiments into lethally irradiated syngeneic hosts and allowed recovery for up to several weeks prior to tamoxifen-inducible KO of Zeb2 or Zeb1 alone. Induced Cre in this system lead to Zeb2 hematopoietic deficient mice that phenocopied the original Zeb2 phenotypes observed using the interferon-inducible Mx1-Cre model. These defects included excessive granulocyte differentiation, decreased monocyte, and B cell differentiation, as well as defects in erythroid and megakaryocytic differentiation, at 8 weeks after inducing the Zeb2 KO [9] (S4 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Double deletion of Zeb1 and Zeb2 causes PB cytopenia and severe differentiation defects in HSPCs.

(A) Schematic of experiments (left panel) used to study the effect of tamoxifen-inducible deletion of Zeb1, Zeb2, or both after donor BM reconstitution (CD45.2⁺) of lethally irradiated recipients (CD45.1⁺). Schematic of loxP flanked (floxed-Fl) conditional Zeb1 and Zeb2 alleles before tamoxifen induced Cre-mediated deletion and recombined delta Zeb1 and Zeb2 alleles after recombination (right panel). (B) HCT analysis of BM 10 days after tamoxifen treatment showing decreased HGB (top left) in Zeb1^Δ/Δ, Zeb2^Δ/Δ, and Zeb1/2^Δ/Δ DKO settings. There was as well decreased RBC (top middle), PLT (top right), and GRA (bottom left). LYM (bottom middle) in the BM were decreased only in the Zeb1/2^Δ/Δ DKO. (C) Flow cytometric analysis of HSPCs in the BM 10 days after tamoxifen treatment showing increased numbers of LSK cells and well as LT-HSCs (LTS- lin⁻cKit⁺Sca1⁺CD34⁻Cd125⁻) in Zeb1/2^Δ/Δ DKO settings. (D) Representative flow cytometry analysis for hematopoietic stem and progenitor populations showing increased numbers of LSK and LT-HSCs in Zeb1/2^Δ/Δ DKO settings. Bars in panels represent mean ± SD, n = 5 per group; *p < 0.05; **p < 0.01; ****p < 0.0001, Dunnett multiple comparisons test. Raw data behind graphs are included in C of S1 Data. BM, bone marrow; DKO, double knockout; GRA, granulocyte; HCT, hematocrit; HGB, hemoglobin; HSPC, hematopoietic stem and progenitor cell; LSK, Lin⁻Sca1⁺cKit⁺; LT-HSC, long-term HSC; LTS, long-term HSC; LYM, lymphocyte; MPP, multipotent progenitor; PB, peripheral blood; PLT, platelet; RBC, red blood cell; STS, short-term HSC.

https://doi.org/10.1371/journal.pbio.3001394.g004

We performed similar Zeb1 KO experiments with the Rosa26-Cre-ERT2 line (S5A Fig) and could confirm efficient inactivation of the conditional Zeb1 allele in whole BM cell populations (S5B Fig). Moreover, we confirmed the cell-autonomous defects of Zeb1-deficient HSPCs in methylcellulose replating experiments that we observed using the constitutive Vav-iCre model. Further, we demonstrated that tamoxifen-mediated KO of Zeb1 in HSPCs severely compromised their ability to give rise to hematopoietic colonies that was already present at initial plating. However, this became sequentially worse in terms of number of colonies generated (compared to controls) by the second replating experiments (S5C Fig: **p < 0.01). Finally, analysis of T-cell contribution of Zeb1-deficient adult HSPCs from this inducible model also showed similar T-cell defects as those documented in the constitutive Vav-iCre model, with decreased DN4⁺ cells, but, here, we observed increased skewing to CD4⁺ T-cell development (S5D and S5E Fig: *p < 0.05, ***p < 0.001).

These results demonstrate that the inducible Rosa26-Cre-ERT2 line can be used in a similar manner as the Mx1-Cre line in the context of BM transfer settings to efficiently examine the cell-autonomous roles of Zeb1/2 in long-term hematopoiesis settings.

To determine the consequences of simultaneous KO of both Zeb1 and Zeb2 on steady-state hematopoiesis, similar BM transplants were performed using R26-CreERT2 control, single R26-Cre-ERT2; Zeb1 ^fl/fl (Zeb1^Δ/Δ), R26-Cre-ERT2; Zeb2 ^fl/fl (Zeb2^Δ/Δ), and iDKO R26-Cre-ERT2; Zeb1 ^fl/fl; Zeb2 ^fl/fl (Zeb1/2^Δ/Δ) BM into lethally irradiated syngeneic recipients. Following recovery from irradiation, recipients were given tamoxifen via oral gavage for 3 consecutive days (Fig 4A). Within 2 weeks of the final dose of tamoxifen, the Zeb1/2^Δ/Δ mice had to be euthanized due to anemia, severe weight loss, and general ill health (S6A Fig). Immediate early changes in both percentages of live cells and absolute cell numbers of mature blood cells 10 days after tamoxifen showed extreme fluctuations in blood cell composition in the PB in single Zeb1-, Zeb2-, and iDKOs compared to R26-Cre-ERT2-only controls (S6B Fig: *p < 0.05, **p < 0.01, ****p < 0.0001). BM analysis at this time point showed consistently decreased hemoglobin (HB) levels in all 3 mutant backgrounds, with more severe effects observed in both the Zeb2^Δ/Δ and Zeb1/2^Δ/Δ samples (Fig 4B, top left panel). Similar decreases were also observed in red blood cell (RBC) numbers (Fig 4B, top middle panel). Platelet (PLT) numbers were also decreased in all 3 mutant backgrounds, but the most severe decreases were observed with Zeb1/2^Δ/Δ samples (Fig 4B, top right panel; *p < 0.05, ***p < 0.001, ****p < 0.0001). Similar defects in the numbers of granulocytes were observed, again with the most severe decreases in the Zeb1/2^Δ/Δ mice (Fig 4B, lower left panel: *p < 0.05, **p < 0.01, ***p < 0.001). Significant decreases in lymphocyte (LYM) numbers were only observed in the Zeb1/2^Δ/Δ samples at this early time of analysis (Fig 4B, lower right panel: **p < 0.01).

Flow cytometric analysis of HSPC populations from the BM showed a significant increase in the percentage of live LSK and long-term hematopoietic stem cells (LT-HSC-LSK-Cd135⁻CD34⁻) per total number of cells in the femur (Fig 4C and 4D, S6C Fig: ****p < 0.0001). This is suggestive of a more severe block in multilineage differentiation associated with the inducible loss of both Zeb1/2 than is observed in either Zeb2 or Zeb1 single mutants.

Overall, these results suggest that inducible KO of Zeb1 and Zeb2 in the adult hematopoietic system can lead to rapid and lethal decreases in RBC and PLT numbers. These are associated with very early blocks in LT-HSC differentiation that are more severe than genetic inactivation of either Zeb2 or Zeb1 alone.

Continued presence of a single Zeb2 allele rescues iDKO hematopoietic differentiation

During the establishment of the complex Rosa26-Cre-ERT2; Zeb1 ^fl/fl; Zeb2 ^fl/fl triple transgenic background, we also established cohorts of Rosa26-Cre-ERT2; Zeb1 ^fl/fl; Zeb2^fl/+ mice with a single wt Zeb2 allele (Zeb2^Δ/+), which we included in our subsequent hematopoietic analysis, 10 days after tamoxifen administration. Such Zeb2^Δ/+ mice did not show any evidence of severe weight loss or anemia. Consistent with this finding, hematocrit (HCT) analysis of mice that carried this allele rescued the HCT defects observed in mice, including restoration of RBC, HB, and PLT cell percentages (Fig 5A: *p < 0.05, **p < 0.01, ****p < 0.0001). They also have normalized overall white blood cell (WBC)/LYM and NEU numbers in the PB (S6D Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Maintenance of a single Zeb2 wt allele rescues PB cytopenia and severe differentiation defects in Zeb1/2 DKO HSPCs.

(A) Hemavet analysis showing that maintenance of a single wt Zeb2 allele in Zeb2^Δ/+, Zeb1Δ/Δ (abbreviated Zeb2^Δ/+) mice can rescue the HCT defects observed in Zeb1/2^Δ/Δ DKO including RBC and HB levels as well as normalization of PLT (right) numbers. (B) Flow cytometric analysis of HSPCs from the BM showing that presence of single Zeb2 allele in Zeb2^Δ/+ mice can also normalize LSK, LT-HSC (lin⁻cKit⁺Sca1⁺CD34⁻Cd135⁻), ST-HSC (lin⁻cKit⁺Sca1⁺Cd34⁺Cd135⁻), HPC (lin⁻cKit⁺Sca1⁻), CMP, GMP, and MEP HSPC numbers compared to defects observed in Zeb1/2^Δ/Δ DKOs. (C) Representative flow cytometry plot of HSPCs. Bars in panels represent mean ± SD, n = 5 per group; *p < 0.05; **p < 0.01; ****p < 0.0001, Dunnett multiple comparisons test. Raw data behind graphs are included in D of S1 Data. DKO, double knockout; HB, hemoglobin; HCT, hematocrit; HSPC, hematopoietic stem and progenitor cell; PB, peripheral blood; PLT, platelet; RBC, red blood cell; wt, wild-type.

https://doi.org/10.1371/journal.pbio.3001394.g005

Flow cytometric analysis of the HSPC compartment demonstrated that the presence of a single wt Zeb2 allele also could normalize the increased percentage of LSK, LT/ST-HSCs present in Zeb1/2^Δ/Δ mice as well as normalize HPC populations, including GMP and MEPs (Fig 5B and 5C, S6E Fig: *p < 0.05, **p < 0.01, ****p < 0.0001).

Overall, this early analysis of the respective inducible Zeb1, Zeb2 and Zeb1/2 KOs as well as the presence of a single Zeb2 wt allele suggests that Zeb2 and Zeb1 can cooperate in controlling hematopoietic lineage fidelity. Further, a single allele of Zeb2 is enough to normalize these severe defects in hematopoiesis observed in Zeb1/2 DKOs.

RNA-seq analysis of LSK-enriched populations reveals both common and unique immediate early gene expression programs controlled by Zeb1 and Zeb2

We performed FACS to isolate LSK cells from the BM of Zeb1^Δ/Δ, Zeb2^Δ/Δ, and double Zeb1/2^Δ/Δ KO mice, as well as mice carrying a single Zeb2 wt allele (Zeb1^Δ/Δ; Zeb2^Δ/+). LSK cells were harvested 24 hours after the last administered dose of tamoxifen (4 days after tamoxifen treatment was initiated), and RNA was isolated to identify immediate early differentially expressed genes (DEGs) between these 4 separate genotypes by RNA-seq (N = 3/genotype). This time point was also chosen given the severe lethal phenotype observed in Zeb1/2^Δ/Δ mice 10 days post-tamoxifen. Using the RNA-seq data obtained from Zeb1^Δ/Δ, Zeb2^Δ/Δ, and Zeb1/2^Δ/Δ KO LSK cells and bioinformatics analysis, we were able to quantify the degree to which Exon 6 of Zeb1 and Exon 7 of Zeb2 were deleted and the residual amount of unrecombined conditional alleles of Zeb1 and Zeb2 that remained at this early time point (S7 Fig). Here, we could demonstrate that 4 days of tamoxifen treatment induced approximately 60% loss of Exon 6 in Zeb1^Δ/Δ and Zeb1/2^Δ/Δ KO cells compared to Zeb2^Δ/Δ LSK cells (S7A Fig). Comparing Zeb2^Δ/Δ and Zeb1/2^Δ/Δ KO cells with Zeb1^Δ/Δ LSK cells showed an approximate 90% loss of Exon 7 containing transcripts for Zeb2 (S7B Fig).

Principle component analysis (PCA) of DEGs showed good clustering among samples with the same genotype, with Zeb1^Δ/Δ and single Zeb2 wt allele samples clustering closer together, whereas Zeb2^Δ/Δ and Zeb1/2^Δ/Δ samples clustered further apart, representing the fact that these are more genetically divergent (S8A Fig). Comparing overall DEG heatmaps from Zeb1^Δ/Δ and the 3 other generated genotypes using Degust bioinformatics tools using Zeb1^Δ/Δ samples for the respective comparisons, we found many genes up-regulated in Zeb2^Δ/Δ LSK cells that are further up-regulated in Zeb1/2^Δ/Δ DKOs. A vast majority of these genes are re-repressed, or their normal levels maintained in the single Zeb2 wt allele “rescue” (S8B Fig). These results would imply that Zeb1 and Zeb2 can act together to repress specific gene expression programs involved in hematopoietic differentiation. Volcano plots comparing DEGs between Zeb1^Δ/Δ and Zeb1/2^Δ/Δ samples show the expected decreased expression in Zeb2 in Zeb1/2^Δ/Δ samples that are null for both Zeb2 and Zeb1, compared with Zeb1^Δ/Δ LSKs that are null for Zeb1 and are wt for Zeb2.

Moreover, further validation of this comparative approach is that some of the top DEGs that are overexpressed in the Zeb1/2^Δ/Δ samples include Ctse, Id2, and Epcam that have been previously identified as being transcriptional targets of Zeb2 that are normally repressed by Zeb2 in T- [29], dendritic [30], and macrophage [31] lineage cells, respectively (S8C Fig). These targets among several others are significantly re-repressed or maintained in Zeb2 single wt allele LSK samples compared with DKO Zeb1/2^Δ/Δ samples (S8D Fig). These maintained genes include genes involved hematopoietic adhesion, homing, and niche modulation (including LSR, Ccl6, Itga1, Ccr9, Ccr5, Ccr2, and Cxcr3), stemness genes (Slamf7, Lgals3, Epcam, and Alcam), transcriptional regulators (Bcl6, Id2, and mycl), and signaling (Tgfbi, Fgfr1, Acvrl1, and Il13ra1) (S8E Fig).

To identify DEGs relating to immediate early ZEB2-dependent genes more robustly, we have used a bioinformatics approach using a stringent false discovery rate (FDR) <0.1 cutoff [32]. Here, we have compared the DEGs between Zeb1^Δ/Δ and Zeb1/2^Δ/Δ (Zeb2 DEGs), Zeb1^Δ/Δ; Zeb2^Δ/+ versus Zeb1/2^Δ/Δ samples (Zeb2 DEGs rescue), and Zeb2^Δ/Δ versus Zeb1/2^Δ/Δ (Zeb1 DEGs) (Fig 6A). These DEGs between these 4 separate genotypes are depicted in a Z-score heatmap (Fig 6B). Merging the first 2 datasets, we have identified 143 common DEGs from these 2 Zeb2 genetic settings (total of 321 Zeb2 DEGs, S2 Table) (Fig 6C left) and 77 Zeb1 DEGs, of which 32 are shared with Zeb2 (Fig 6C right, S2 Table). Although these numbers of differentially regulated genes are relatively low, they possibly represent some of the first immediate early and possibly direct transcriptional targets of ZEB1/2 that are affected in LSK cells, upon the inducible loss of Zeb1 or Zeb2 alone or in Zeb1/2 DKOs. What is most noticeable again from the heatmap in Fig 6B is a vast majority of DEGs being up-regulated in Zeb1/2^Δ/Δ settings implying a strong degree of cooperation or synergy between ZEB1 and ZEB2 in gene repression (rightmost column in Fig 6B). Moreover, there is a subset of DEGs that appear to be up-regulated specifically in Zeb1^Δ/Δ (top leftmost panel in Fig 6B), but as well appear to be normalized in Zeb2^Δ/+ versus Zeb1/2^Δ/Δ samples when one wt Zeb2 allele is maintained.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. RNA-seq analysis of LSK-enriched populations reveals both common and unique immediate early gene expression programs controlled by Zeb1 and Zeb2.

(A) DEG lists obtained with edgeRun R package, when comparing R26-Cre-ERT2; Zeb2^fl/fl (Zeb2^Δ/Δ), R26-Cre-ERT2; Zeb2^fl/+, Zeb1^fl/fl (Zeb2^Δ/+, Zeb1^Δ/Δ), and Zeb1^fl/fl (Zeb1^Δ/Δ) against iDKO R26-Cre-ERT2 Zeb1/2 ^fl/fl (Zeb1/2^Δ/Δ), respectively. To define DEGs, we used as cutoff an FDR <0.1. (B) Heatmap of >360 combined ZEB1 and ZEB2 DEGs, sorted from most induced to repressed ZEB2-DEGs. From left to right, we plotted the Z-scores of gene expression of R26-Cre-ERT2 Zeb2 ^fl/fl (Zeb2^Δ/Δ), single R26-Cre-ERT2; Zeb2^fl/+ allele, Zeb1^fl/fl (Zeb2^Δ/+, Zeb1^Δ/Δ) and the iDKO R26-Cre-ERT2; Zeb1/2^fl/fl (Zeb1/2^Δ/Δ), respectively. (C) (Left) Intersections of DEGs from Zeb1 null cells expressing double or a single Zeb2 allele against iDKO cells, respectively. (Right) Intersections of DEGs from Zeb1 or Zeb2 null cells against iDKO cells, respectively. Raw data behind (B) are included in S2 Table. DEG, differentially expressed gene; DKO, double knockout; FDR, false discovery rate; iDKO, inducible double knockout; LSK, Lin⁻Sca1⁺cKit⁺; RNA-seq, RNA sequencing.

https://doi.org/10.1371/journal.pbio.3001394.g006

Gene ontology (GO) analysis using several sources such as BioCarta, BioCyc, GO, KEGG, and Reactome as queries [33] for ZEB2 differential targets genes identified inappropriate gene activation in LSK cells of genes that should normally only be expressed in more mature immune cells such as those involved in MHC class II antigen presentation, leukocyte migration, or processes reserved T-cell functions including T-cell migration and chemotaxis (Fig 7A). Associated network analysis using the STRING database highlighted that many of the DEGs exist in gene regulatory networks controlling chemotaxis as well as antigen processing and presentation via MHC class II that is linked to antimicrobial responses and Toll-like receptor signaling (Fig 7B, S3 Table). Analysis of intact ZEB1-dependent DEGs shows evidence of aberrant gene regulation involved in mitochondrial functions including respiratory chain complex and NADH dehydrogenase activity (Fig 7C and 7D, S3 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. RNA-seq analysis of LSK-enriched populations reveals both common and unique immediate early gene expression programs controlled by ZEB1 and ZEB2.

(A) Associated GO terms with ZEB2-DEGs, obtained with the STRING database, using an FDR <0.05). (B) Associated network analysis for ZEB2-DEGs, obtained with STRING database using highest confidence interaction scores (0.900) and clustered with an MCL inflation parameter of 3. (C) Same as (A) for ZEB1 DEGs. (D) Same as (B) for Zeb1 DEGs. Raw data behind panels are included in S3 Table. DEG, differentially expressed gene; FDR, false discovery rate; GO, gene ontology; LSK, Lin⁻Sca1⁺cKit⁺; RNA-seq, RNA sequencing.

https://doi.org/10.1371/journal.pbio.3001394.g007

Using the strict criteria described, only 32 genes were designated as being “shared” between Zeb2-DEGs and Zeb1-DEGs, but this may be significantly underestimated because of the stringent analysis performed and the lack of comparison with Cre and tamoxifen-only controls. Of these 32, several genes (mt-Nd3, mt-Nd4l) involved in mitochondrial metabolism have emerged, as well as other genes such as Epcam and Gprc5c involved in cancer stemness and dormancy of HSCs, respectively [34,35], and Cdcp1 that is expressed on leukemic blasts [36]. The common DEG CCl6 has been implicated in impaired HSC homeostasis [37]. Other interesting Zeb2-only DEGs include Tgfbi and Alcam that play important roles in HSC biology [38,39] and Id2 that can significantly affect hematopoietic lineage priming [40]. Csf1r is another DEG that plays essential roles in monocyte development [41]. Interestingly, the Zeb1-only DEG Mat2a, an important determinant in MLL-driven AML, is up-regulated in Zeb1 null settings [42]. Moreover, Tdgf1 (Cripto), Neo1 (Neogenin-1), and Mapk13 that play roles in HSC function and stress responses were increased in Zeb1-deficient settings [43–45]. Lastly, altered expression of Vwf (up) and Ube2t (down) genes that are involved in clotting/vascular disorders and Fanconi anemia, respectively, were also DEGs in Zeb1^Δ/Δ LSKs [46,47].

To verify if these DEGs identified by RNA-seq can also be directly controlled at the chromatin level by ZEB1 and ZEB2, we reanalyzed publicly available Chromatin Immunoprecipitation Sequencing (ChIP-seq) datasets of ZEB1 and ZEB2 in GM12878 and K562 human cell lines, respectively, plotting the occupancy of these factors in human genes homolog to the DEGs in mouse (see S4 Table). Although these 2 cell lines have different hematopoietic origins, we observed marked occupancy of both transcription factors around the gene start, with most of the peaks spanning 0 to 1 kb around these sites (S9A and S9B Fig). Moreover, ZEB1 and ZEB2 shared a substantial proportion of peaks and some genes contained ZEB1 and ZEB2 binding enrichment in the same genomic regions (S9C Fig). Thus, many genes controlled by Zeb1 and Zeb2 in mouse are also potential transcriptional targets bound by ZEB1/2 in human cells, demonstrating the existence of a conserved network controlled by ZEB1/2.

Overall, we have discovered that many genes may be co-repressed by ZEB1 and ZEB2 that become highly expressed in DKO settings. This finding fits well with the generally accepted view that both ZEB1 and ZEB2 are thought to be predominantly transcriptional repressors in the context of LSK cells and serve to repress genetic programs that should only be activated in more mature lineage-specified cells. Moreover, specific genes were identified that may be associated with altered HSC homeostasis and lineage-specific differentiation/function and are discussed in further detail below.

Zeb1 and Zeb2 overexpression both drive extramedullary hematopoiesis and monocyte skewing

Previous studies have demonstrated that ZEB1 and ZEB2 may play oncogenic roles in myeloid cell transformation particularly in MLL-AF9 driven models [16,17]. Work presented here would suggest that loss of ZEB1 can interfere with both monocytic and neutrophil differentiation, whereas ZEB2 loss can lead to enhanced granulocytic differentiation in addition to defects (like ZEB1) in monocytic/macrophage differentiation.

In order to examine the effects of ZEB1 and ZEB2 gain of function on hematopoiesis, we have intercrossed conditional Rosa26 locus Zeb1 [48] and Zeb2 [15] cDNA-based alleles with the hematopoietic-restricted Vav-iCre/Tie2-Cre lines. Ubiquitous hematopoietic Rosa26 locus-based overexpression (3- to 4-fold as mRNA) of Zeb1 (Fig 8B; p = 0.0125) along with EGFP expression (Fig 8C) and Zeb2 (S10A Fig) in both models led to enlarged spleen sizes already by 3 to 4 months of age that is indicative of extramedullary hematopoiesis (Fig 8E, S10B Fig: *p < 0.05). Furthermore, myeloid lineage expansion was observed in both ZEB1 and ZEB2 models in BM, spleen, and PB as judged by increased numbers of CD11b⁺, Gr1⁺ cells (Fig 8F and 8G, S10C and S10D Fig: *p < 0.05, **p < 0.01). Using further Ly6G staining, we confirmed that increased myeloid compartment consisted mainly of CD11b⁺, Ly6G⁻ monocytic lineage cells in the BM of both Vav-iCre Rosa26-Zeb1^tg/tg and Tie2-Cre driven Rosa26-Zeb2^tg/tg models (Fig 8H, left and right, respectively; S10E Fig: *p < 0.05, **p < 0.01).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Zeb1 overexpression leads to extramedullary hematopoiesis/splenomegaly, enhanced myeloid cell development, and monocyte lineage skewing.

(A) Schematic of conditional Rosa26-Zeb1-IRES-EGFP-pA+ transgenic locus (left). (B) Following Vav-iCre-mediated deletion of the loxP flanked transcriptional stop cassette Zeb1 expression is increased approximately 4.5-fold in transgenic BM HSPCs compared to controls (N = three 5-month-old female mice/genotype, p = 0.0125; error bars indicate SD of the mean, Mann–Whitney test) along with (C) dosage-dependent EGFP expression in both heterozygous and homozygous Zeb1 transgenic HSPCs (Flow cytometry for EGFP). (D) Western blot confirmation of increased ZEB1 protein in the BM and spleen of Vav-iCre; Zeb1^tg/tg mice compared to Cre-negative control samples. (E) Increased spleen size/extramedullary hematopoiesis seen in Zeb1^tg/tg transgenic mice (left panel) showing roughly doubling in size compared to body weight (right panel). (F) Flow cytometric analysis showing increased myeloid cells in spleen (CD11b⁺, Gr1⁺). (G) Representative flow cytometry plot showing increased CD11b⁺, Lys6G^- monocytes in the BM of Zeb1^tg/tg mice. (H) Increases in myeloid cells (CD11b⁺, Gr1⁺) with monocytic skewing/expansion was present in both Vav-iCre (left) and Tie2-Cre models (right). Data are represented as mean + SD from 3 biological replicates/genotype. *p < 0.05; **p < 0.01, nonparametric t test. Raw data behind graphs and western blot in (D) are included in E of S1 and S2 Data, respectively. BM, bone marrow; HSPC, hematopoietic stem and progenitor cell; SP, spleen.

https://doi.org/10.1371/journal.pbio.3001394.g008

Overall, these results are consistent with the fact that Zeb1 and Zeb2 levels are both essential for monocytic differentiation and can also drive monocytic lineage differentiation when their expression levels increase but are more divergent in their roles in granulocytic lineage differentiation.

Of note, hematopoietic-restricted expression of Zeb2 does lead to the development of ETP-ALL from 6 months of age onward [15] and only leads to AML development when intercrossed onto a p53 conditional null tumor-prone background [15]. Hematopoietic Zeb1 overexpressing mice do not spontaneously develop T-cell or myeloid malignancies up to 1.5 years of age and appear to have a normal life expectancy [48].

Lack of synergy between Zeb2 and Zeb1 loss in influencing MLL-AF9 driven leukemic progression

Previous work has determined that ZEB2 is a key genetic determinant in AML initiation/progression and that Zeb2 knockdown (KD) in retroviral MLL-AF9 models of AML decreases cellular proliferation and enhances myeloid differentiation in vitro [16]. Using novel inducible mouse models of MLL-AF9 myeloid transformation, Zeb1 KD was found to decrease proliferation, increase adhesion, and decrease migratory properties of AML cells in vitro and decreased cell infiltration of AML cells into the BM and other organs in vivo [17]. These results together with their common detrimental effects on myelopoiesis would suggest potential synergy of Zeb1/2 overexpression in AML progression/maintenance.

To examine the role of Zeb2 in AML progression in vivo, we have used Rosa26-CreERT2; Zeb2 ^fl/fl BM and an MLL-AF9 retrovirus to create primary AML leukemic cells that allowed us the ability to temporally inactivate Zeb2 in vivo in a tamoxifen-inducible manner after secondary engraftment (Fig 9A). Here, we could demonstrate that inducible Zeb2 inactivation could significantly enhance the survival of mice transduced with MLL-AF9 secondary leukemic cells compared to Cre-only controls (Fig 9B; p = 0.0018). In a separate set of experiments tamoxifen-inducible KO of both Zeb1 and Zeb2 (although significant) did not extend the survival of mice transplanted with secondary MLL-AF9 AML (Fig 9C; p = 0.0031) over Zeb2 inactivation alone and if anything, decreased the overall survival between the 2 groups (70 days for Zeb2 KO alone versus 58 days median survival for Zeb1/2 DKO; Fig 9B and 9C). Of note, we documented significant effects of induced Cre activity alone on enhancing survival that is probably related to activation of DNA damage response pathways elicited by Cre binding to pseudo loxP sites in the genome [49,50]. Moreover, there were differences in the onset of lethality in the Cre-only + tamoxifen controls in the 2 groups (30 days for Cre alone + tamoxifen versus 42 days median survival for Cre alone + tamoxifen in the second set of experiments; Fig 9B and 9C). This may potentially reflect differences in the numbers of tumor-initiating cells present and/or DNA damage response in the individual cell populations between the separate experiments.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Inducible deletion of Zeb1/2 increases in vivo survival in MLL-AF9 secondary transplant settings.

(A) Schematic of inducible deletion strategy to investigate the effects of Zeb2 and Zeb1/2 deletion on secondary leukemia progression. (B) Tam-induced Zeb2 deletion was found to significantly increase overall survival of mice transplanted with MLL-AF9 secondary tumor cells compared to nontreated Veh treated controls (median survival 66 versus 42 days, p = 0.0082, Mantel–Cox test). There was a significant effect of Tam in Cre only treated samples compared to Veh controls (median survival 29 versus 19 days, p = 0.0082, Mantel–Cox test). (C) Tam-induced deletion of both Zeb1 and Zeb2 was also found to significantly increase overall survival of mice transplanted with MLL-AF9 secondary tumor cells compared to nontreated Veh treated controls (median survival 58 versus 31 days, p = 0.0031, Mantel–Cox test). There was a significant effect of Tam in Cre only treated samples compared to Veh controls (median survival 42 versus 31 days, p = 0042, Mantel–Cox test). N = 5 mice/treatment group for all arms of the experiments. Raw data behind graphs are included in F of S1 Data. BM, bone marrow; PB, peripheral blood; Tam, tamoxifen; Veh, vehicle.

https://doi.org/10.1371/journal.pbio.3001394.g009

We also reviewed the expression levels of ZEB1 and ZEB2 across AML samples available in cBioPortal database [51]. As evidenced in mouse, the levels of ZEB1 are in general statistically lower than ZEB2 in primary AML cells, with the translocation driver mutation likely affecting the expression of these transcription factors (S11A Fig, left and right, respectively: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). Kaplan–Meier survival analysis indicate that overall higher levels of ZEB1 are associated with poor prognosis and decreased overall survival irrespective of translocation, supporting an oncogenic role of ZEB1 (p = 0.048 Bonferroni adjusted t test, see S11B Fig, left and right, respectively). Conversely, ZEB2 levels did not appear to statistically influence prognosis in human AML (p = 1.0 Bonferroni adjusted t test) with regard to overall survival. Further analysis of ZEB1 and ZEB2 mRNA expression demonstrates that the encoded transcription factors are significantly higher in the leukemic blast population than in the bulk tumor population (S11C Fig: p = 0.0417 and p = 0.0024, respectively). Therefore, expression of the mRNAs for these transcription factors may be diluted in the bulk RNA-sample analysis (S11A Fig) as well as in overall levels of expression used in the Kaplan–Meier survival curves (S11B Fig).

Note added in proof

During the final preparation of this manuscript, Almotiri and colleagues published a similar study [63] as ours using the same conditional Zeb1 allele that we have used in this study [24]. However, their analysis of ZEB1 predominantly used the inducible Mx1-Cre mouse model to delete Zeb1 alone, whereas this study predominantly used constitutive hematopoietic enhanced Tie2 and restricted Vav-iCre lines with BM transplants to determine the cell-autonomous role of Zeb1 in hematopoiesis. In general, there is a significant agreement between these 2 studies especially pertaining to the monocyte defects and loss of reconstitution potential of Zeb1-deficient HSPCs. The focus of their manuscript is more on T-cell defects that were previously reported in the full Zeb1 null study [11], whereas we focused more on the novel synergistic roles between ZEB1 and ZEB2 in maintaining hematopoietic lineage fidelity. Subtle differences in the results between the 2 papers may originate from the other groups using the Mx1-Cre model, nonhematopoietic expression of Cre in the stroma, as well as potential effects of poly:IC and its known interferon response and potential effects of HSC quiescence [64]. Fetal liver derived HSCs used in our study are more highly proliferative than BM-derived adult HSCs and appear to utilize oxidative metabolic pathways more than BM HSCs and may be better protected from reactive oxygen species (ROS)-mediated genotoxicity [65]. Moreover, human fetal and adult HSCs have been demonstrated to give to distinct T-cell lineages in humans [66]. Given that we do see differences in T-cell differentiation of Zeb1-deficient HSPCs in our study compared to Almotiri and colleagues, this may be a contributing factor.

One area where there is some discrepancy between the 2 studies is that Almotiri and colleagues suggest that Zeb1 plays a tumor suppressor role in AML progression whose loss enhances AML lethality. In the present study, we have investigated potential synergy between Zeb1 and Zeb2 loss, and, although loss of Zeb1 did decrease overall survival slightly in Zeb1/2 DKO MLL-AF-9 settings compared to Zeb2 loss alone, there was still significant survival advantage observed in Zeb1/2 DKOs compared to Cre-only controls. There was as well significant variability observed in these studies with clear effects of Cre-alone, affecting the outcomes.

Our work with Rosa26-based transgenic overexpression models would suggest equal detrimental effects of Zeb1 overexpression as Zeb2 on myeloid differentiation, and our work would tend to support previous reports that Zeb1 can act as an oncogene in certain forms of AML [11]. Additionally, increased levels of ZEB1 mRNA are associated with decreased overall survival of AML patients that would be consistent with an oncogenic role. It is clear from our work and that of others that the concept of oncogene and tumor suppressor as it pertains to ZEB1 and ZEB2 is very contextual and depends on cell type examined and stage of cell differentiation when Zeb1/2 mRNA levels become dysregulated [17]. Moreover, from our analysis and the work from Stavropoulou and colleagues [17] elevated expression of Zeb1 and potentially Zeb2 transcripts in the leukemic stem cell compartment may ultimately drive poor outcomes as opposed to overall expression of Zeb1/2 in bulk AML samples.

Animal experimentation and handling

The Tie2-Cre [25], tamoxifen (Tam)-inducible ROSA26^CreERT2/+ mice [27], and conditional Zeb2-KO [28], Zeb1-KO [24] mouse model have been backcrossed to a C57Bl/6 genetic background for at least 10 generations. All strains were bred in-house in specific pathogen-free (SPF) facility.

E14.5 cells from an entire fetal liver from Tie2-Cre Zeb1^+/+, Tie2-Cre Zeb1 ^fl/fl CD45.2⁺ backgrounds were transplanted via tail vein injection into lethally irradiated (2 × 550 Rads) CD45.1⁺ hosts (1 fetal liver/host).

Competitive BM experiments were performed using equal numbers (2.5 × 10⁶) of CD45.2⁺ Vav-iCre, Zeb1 ^fl/fl, Vav-iCre Zeb1^fl/+, or Vav-iCre only doner cells and CD45.1+ competitor cells that were transplanted by tail vein injection into lethally (2 × 550 Rads) irradiated CD45.1 hosts.

For inducible deletion experiments, 5.0 × 10⁶ BM donor cells were used from CD45.2⁺ ROSA26^CreERT2/+ Zeb1^+/+ Zeb2^+/+, ROSA26^CreERT2/+ Zeb2^fl/fl, ROSA26^CreERT2/+ Zeb1^fl/fl, ROSA26^CreERT2/+ Zeb2^fl/+ Zeb1^fl/fl, and ROSA26^CreERT2/+ Zeb1/2^fl/fl backgrounds to reconstitute lethally irradiated (2 × 550 Rads) CD45.1 recipients. These mice were allowed to recover for between 2 and 6 weeks before they were orally gavaged for 3 consecutive days with tamoxifen (5 mg in 50-ul vehicle per 25-g body weight).

In MLL-AF9 initiation and progression experiments, MLL-AF9 transduced C57BL/6 (CD45.2) fetal liver cells (ROSA26^CreERT2/+, Zeb1^+/+, Zeb2^+/+, ROSA26^CreERT2/+ Zeb2^fl/fl or ROSA26^CreERT2/+, Zeb1^fl/fl, Zeb2^fl/fl) were collected under aseptic conditions and were intravenously injected into lethally irradiated (2 × 550 Rads) C57BL/6 (CD45.1) recipient mice. For secondary leukemia transplantation studies, 10,000 GFP⁺ckit^hiCD11b⁺ primary leukemia cells were intravenously injected into sublethally irradiated (550 Rad) recipient mice. Inducible deletion of the Zeb gene was achieved following exposure of CreERT2, Zeb1^fl/fl, Zeb2^fl/fl mice to tamoxifen (5-mg tamoxifen in 50-ul vehicle per 25 g) by oral gavage once daily for 3 consecutive days. All irradiate mice were maintained on acidified water following irradiation.

All experiments were performed according to the regulations and guidelines of the Ethics Committee for care and use of laboratory animals of Monash University (E/1690/2016/M, #5789) and the University of Manitoba (#18–050).

Flow cytometric and HCT analysis

Cells were stained with antibodies listed in B in S5 Table according to the manufacturer guidelines. Flow cytometric analyses were performed on the LSRII and Fortessa X-20 cytometer (BD Biosciences, Sydney, Australia), and the results were analyzed by FACSDiva or FlowJo software (BD Biosciences). Cells for MLL-AF9 experiments and RNA-seq were stained and sorted on Influx or FACSAria Fusion sorters (BD Biosciences) at AMREP Flow Cytometry Core Facility and FlowCore, Monash University. All flow cytometry data generated and/or analyzed during the current study are available on Zenodo (doi: 10.5281/zenodo.5498282).

Submandibular blood samples were collected into EDTA-coated tubes, and hematology parameters were measured using a HemaVet 950FS automated blood analysis machine (Drew Scientific, Miami Lakes, Florida, USA).

Methylcellulose culture and replating

E14.5 fetal livers were isolated and cultured in methylcellulose 3434 (Stem Cell Technologies, Vancouver, British Columbia, Canada) for 7 days at 37°C and 5% CO₂ in duplicate. For replating assays, cells were collected and resuspended in 1X PBS and then replated in Methocult 3434 and cultured for another 14 to 21 days.

RNA-seq analysis

Mouse BM cells were stained using antibodies listed in B of S5 Table and LSK cells were sorted into TRIzol Reagent (Thermo Fisher Scientific, Australia). RNA was extracted using Direct-zol RNA microprep kit (Zymo Research, distributed by Integrated Sciences, Chatswood, NSW, Australia) and quality assessed using a BioAnalyzer machine (Agilent Mulgrave, Australia). Library preparation was performed using the Nugen RNA-Seq system V2 (SPIA Amplification) followed by Nugen Ultralow Library, and single-end 75 bp reads were generated on an Illumina NextSeq 500 machine in high output settings. Details of RNA-seq data analysis and exon quantification are included in S1 Text. Raw RNA-seq data files were deposited as a NCBI BIOPROJECT #PRJNA679880.

scRNA-seq data of HSPCs previously published [19] was reanalyzed using pseudotime and trajectory-based analysis of scRNA-seq data [20,21] to gain insight into the levels and temporal expression of Zeb1 and Zeb2 during early hematopoiesis. Details of this analysis are included in S1 Text.

Real-time quantitative PCRs

Total RNA was isolated using RNeasy Plus Mini Kit (Qiagen, Chadstone Centre, Victoria, Australia). cDNA was synthesized using the SuperScript IV First-Strand Synthesis System (Thermo Fisher Scientific, Australia), starting from equal amounts of RNA. qRT-PCRs were performed using the TaqMan Gene Expression Master Mix (Applied Biosystems) on a LichtCycler 480 system (Roche, Australia). Gene expression was standardized against housekeeping genes Gusb, Gapdh, and Hprt. All primers used are listed in part A of S5 Table.

Western blot

Mononuclear cells were purified from the BM and the spleen. Whole cell lysates were prepared by using the RIPA Buffer (10mM Tris-HCl, pH 8.0, 1mM EDTA, 0.5mM EGTA, 1% Triton X-100, 0.1% Sodium Deoxycholate, 0.1% SDS, 140mM NaCl dilute with dH₂O). The proteins were extracted and run by using the 10% gel (Acrylamide/Bis-acrylamide 29:1). The primary antibodies were Rabbit Ab anti-ZEB2 (NBP1-82991), Rabbit ab anti-ZEB1 (NBP1-05987), and Rabbit ab anti-β-actin coupled to HRP(NB600-503SS). Abs were obtained from Novus Biologicals (Victoria, Australia).

Statistical analysis

Data were presented as mean ± SD and indicated in the figures. Comparison between 2 data groups was done by 2-sided Student t test. Dunnett multiple comparisons test was used for statistical analysis between 3 or more experiment groups. Kaplan–Meier survival curves for MLL-AF9 models were generated with GraphPad Prism 7 software and log-rank (Mantel–Cox) test was performed for statistical analysis.