The functional duality of HoxB4 in hematopoietic reconstituting cells



Transplantation of CD34+ hematopoietic reconstituting cells (HRC) is an important treatment modality. The cells needed for clinical hematopoietic reconstitution after myeloablation most commonly derive from bone marrow which have been mobilized into the peripheral blood. The number of CD34+ cells in mobilized blood samples is used to indicate the appropriateness of transplantation although it does not evaluate the two necessary functions for successful transplantation: long-term reconstitution mediated by cells with self-renewing proliferation and short-term hematopoietic differentiation mediated by progenitor cells. Using a novel high-resolution immunophenotyping technology on a flow cytometric platform, we have assessed uniformly mobilized CD34+ cells for expression levels of 16 molecules previously associated with HRC function and sought correlations of these expression data with functional short-term assays for colony formation that do predict successful transplantation. We found that colony-forming units were significantly correlated with HoxB4 expression, which was explained by the number of CD34+ cells in the samples. However, analysis of colony-forming units normalized to the CD34+ cell count revealed a significant negative correlation with HoxB4 expression. Thus, increased levels of HoxB4 enhance CD34+ cell number via self-renewing expansion but concomitantly depreciate the capacity for short-term differentiation per cell. Our findings demonstrate the translation of experimental findings into a clinical setting and suggest that the expression level of HoxB4 in CD34+ cells can be used as a measure of a sample's appropriateness for transplantation. © 2012 International Society for Advancement of Cytometry


As CD34 is expressed on hematopoietic reconstituting cells (HRC), flow cytometric-based enumeration of CD34+ cells has been used as an assay to assess the capacity of a sample to mediate hematopoietic reconstitution. Nevertheless, CD34 expression is not a functional quality biomarker for HRC, as a biological role of the CD34 molecule itself has not been assigned to either the capacity to differentiate into cells of the hematopoietic lineage or with the property of self-renewal. The capacity of HRC preparations to reconstitute the hematopoietic system has been assessed by a granulocyte–monocyte colony-forming unit assay (CFU-GM) and an erythroid burst-forming unit assay (BFU-E); however, these assays measure function and do not give any information at a molecular level.

Many molecules expressed in HRC have been associated with the potential for differentiation into cells of the various hematopoietic lineages and for the capacity for self-renewal including transcription factors (1–14), pathway molecules (15–21), and surface receptors (22–25). Most of these studies have been performed in murine models with knock-out genetic approaches. It has been difficult to assess the expression of these molecules in human HRC because they are expressed at low abundance which precludes quantitative information by standard methods of flow cytometry and because HRC are a small subpopulation of the cells collected making it difficult to determine by western blotting or RT-PCR.

We have used a high-resolution immunophenotyping technology on a flow cytometric platform to assess the expression of molecules associated with HRC function in CD34+ cells from mobilized blood samples obtained from patients with multiple myeloma. Our results demonstrate both positive and negative correlations involving the expression of HoxB4 and assays of HRC function. In addition, correlations among the expression levels of the various molecules assessed were discovered.


Patient Samples

Mobilized blood samples from 52 collections representing 44 unique donors with the diagnosis of multiple myeloma were obtained at University Hospitals Case Medical Center under approval from the Institutional Review Board. HRC were uniformly mobilized from patients with intravenous cyclophosphamide 4 g/m2, filgrastim (Amgen, Thousand Oaks, CA) 10 mcg/kg once or twice daily subcutaneous (determined by resting-state blood CD34+ cell concentration), and prednisone 2 mg/kg/day by mouth for 4 days (26). Apheresis was performed when blood CD34+ cell count exceeded 10 per μl. Filgrastim was continued until the last day of apheresis. A multilumen central venous apheresis catheter was placed either in the internal jugular or subclavian vein for blood cell mobilization and subsequent transplantation. The mononuclear cells were isolated by ficoll/hypaque discontinuous gradient centrifugation and cryopreserved in dimethylsulfoxide for later analysis.


Monoclonal antibodies specific for CD34 and CD117 were obtained from BioLegend (San Diego, CA). Antibodies specific for c-Myc were from Invitrogen (Carlsbad, CA). Antibodies specific for HoxB4 and RUNX1 were from Epitomics (Burlingame, CA). Antibodies specific for GATA-2, Bmi-1, IL-23R, and CXCR4 were from R & D Systems (Minneapolis, MN). Antibodies specific for phosphatase tension homolog (PTEN), phospho-Akt(thr308), phospho-Akt(ser473), and β-catenin were from Cell Signaling Technology (Danvers, MA). Antibodies specific for E47 and IL-3R (CD123) were from BD Biosciences (Mountain View, CA). Antibodies specific for Gab2 and CD130 were from abcam (Cambridge, MA).

Flow Cytometric Analysis

The samples were analyzed for the expression of HoxB4, GATA-2, c-Myc, PTEN, RUNX1, Gab2, phospho-Akt(thr308), phospho-Akt(ser473), β-catenin, Bmi-1, CD117, CD123, IL-23R, CXCR4, and CD130 by Pathfinder Biotech (Cleveland, OH) using enzymatic amplification staining (EAS™) as previously described (27–32). EAS™ is a validated, catalyzed reporter deposition technology based on the enzymatic activity of peroxidase. The events were gated with the characteristic forward scatter and side scatter for CD34+ cells. CD34 counterstaining was included in all samples. In three of the 52 samples obtained, no definitive peak of CD34+ events could be detected; consequently, those samples were not analyzed further. The median fluorescence ratio was obtained for CD34+ events from the median fluorescence intensities for the specific antibodies versus matched control immunoglobulin. Multiple quality control features for high-resolution immunophenotyping have been ascertained. Most importantly, analytical reproducibility was demonstrated by staining identical frozen aliquots of Jurkat cells for a variety of intracellular analytes (31). In addition, we have used carboxylated polystyrene beads substituted with various amounts of human IgG (as an analyte) to demonstrate the linearity of detection by EAS at levels under the level of detection by indirect staining (28). Thus, the data we have obtained are reproducible and quantitative.

Colony-Forming Unit Assays

Mononuclear cells (1 × 105) were grown in duplicate in methylcellulose (Stem Cell Technologies; Vancouver, Canada) containing 10 ng/ml IL-3, 3 U/ml EPO, 50 ng/ml SCF, 10 ng/ml GM-CSF, and Hemin (0.1 mM; Sigma Chemicals; St. Louis, MO), and the cells were incubated at 37°C and 5% CO2. After 12–14 days, colonies greater than 50 cells were identified by morphology, enumerated and expressed as total CFU-GM and BFU-E (33).

Statistical Analysis

The association between two continuous measurements was estimated using Pearson correlation coefficient after checking normality assumption. Logarithmic transformation was performed for those measures whose normality is violated. For each cluster of significant associations identified by Pearson correlation (univariate analysis), the relationships among those measures were further examined by a multiple regression model in which all measures were considered simultaneously. In the model diagnosis, outliers and influential observations were identified using both the studentized residuals and Cook's distance. Statistical analyses were performed using SAS software (Cary, NC). All tests were two-sided and P value less than 0.05 were considered statistically significant.


Enhanced Resolving Power of EAS™ Compared to a Standard Staining Procedure

In previous studies, EAS™ has provided a significantly improved capability of detecting molecular expression levels in flow cytometry (27–32). In this study, we have assessed expression levels of several different analytes in CD34+ cells. The enhancement in specific signals for these analytes in the CD34+ cells is shown in Table 1. The results shown are representative of at least three determinations. It should be noted that in several instances standard staining was unable to resolve expression but EAS was. Even for analytes that showed signals with standard staining, EAS provided a significantly greater dynamic range of detection.

Table 1. Enhancement of Flow Cytometric Detection by EAS™ Compared to Standard Staining Procedures
AnalyteMean fluorescence ratio (Standard staining)Mean fluorescence ratio (EAS™)
  1. CD34+ events were gated after staining with anti-CD34 conjugated with AF-647. The amplified signals were obtained with fluorescein. Results are representative of at least 3 determinations for each analyte.


Detection of Pathway Molecules, Transcriptional Regulators, and Surface Receptors in CD34+ HPC from Mobilized Blood Cells

To begin to elucidate the biology underlying the function of human HRC at a molecular level, we assessed the expression of a variety of pathway molecules (phospho-Akt(thr308), phospho-Akt(ser473), PTEN, β-catenin, and Gab2), transcriptional regulators (HoxB4, c-Myc, GATA-2, E47, RUNX1, and Bmi-1), and surface receptors (CD117, CD130, CXCR4, CD123, and IL-23R) in CD34+ cells from mobilized human peripheral blood samples using high-resolution immunophenotyping that allows for sensitive and quantitative detection of these molecules by flow cytometry (27–32).

Representative results of EAS applied to CD34+ cells for a subset of these analytes are shown in Figure 1. It should be noted that the CD34+ subpopulation of cells could be clearly delineated allowing for a clear assessment of expression in HRC. The expression of the various molecules in HRC was unimodal and discrete, which indicates that expression levels can be precisely and accurately indicated by the calculation of median fluorescence ratios.

Figure 1.

Representative results for the detection of HoxB4, GATA-2, PTEN, and CD117 in HRC. The upper row shows the gate used on the forward-side scatter plot. Mononuclear cells from four different mobilized blood samples were stained for CD34 expression (middle row). The cells were also stained with control immunoglobulin (light outline; lower row) or with specific antibodies (dark outline; lower row) and processed by EAS™ for high resolution immunophenotyping. The amplified signals (lower row) are shown only for the CD34+ events (middle row). Representative results are shown from four different donors to demonstrate the consistency in CD34+ cell delineation. The median fluorescence ratios are shown for the various analytes expressed in the CD34+ cells shown in the lower row. [Color figure can be viewed in the online issue, which is available at]

Correlation of CFU-GM and BFU-E with HoxB4 Expression Levels

Mobilized blood samples were also assayed for CFU-GM and BFU-E, and correlations with the various molecules were sought. As the distribution of BFU-E was skewed, we used logarithmic transformation for this analysis. The expression level of HoxB4 was the only molecule that showed a statistically significant positive association (Fig. 2), and it was correlated with both CFU-GM (r = 0.423; P = 0.003) and logarithm-transformed BFU-E (r = 0.355; P = 0.014). CFU-GM was also marginally significantly correlated with GATA-2 expression levels (r = 0.276; P = 0.06).

Figure 2.

Correlation of HoxB4 expression levels with CFU-GM and log(BFU-E). CD34+ cells from various samples of mobilized blood from patients with multiple myeloma were stained for the expression of HoxB4 and the cells were processed by EAS for high-resolution immunophenotyping. Each sample was also tested for CFU-GM and BFU-E. The distributions of CFU-GM (upper) and log(BFU-E) (lower) versus HoxB4 median fluorescence ratios are shown for the various samples. HoxB4 fluorescence involved a fluorescein-amplified signal. Coefficients of correlation and P values are shown. [Color figure can be viewed in the online issue, which is available at]

The proportion of CD34+ cells among the nucleated blood cells in the mobilized blood samples demonstrated a highly significant correlation with both CFU-GM (r = 0.734; P < 0.0001) and ln(BFU-E) (r = 0.798; P < 0.0001). Moreover, multiple linear regression analysis demonstrated that the proportion of CD34+ cells explains the univariate correlation of HoxB4 with the functional assays.

Besides the association of HoxB4 expression levels on colony forming units by increases in the number of CD34+ cells in the specimens, we were also interested in the effect of HoxB4 per individual cell on short-term hematopoietic differentiation. Consequently, we normalized the number of colonies by the number of input CD34+ cells in the assay and assessed the association of the number of colonies per CD34+ cells in the culture with HoxB4 expression levels (Fig. 3). Although we had shown a positive correlation between HoxB4 and CFU-GM and between HoxB4 and ln(BFU-E), there were statistically significant negative correlations between HoxB4 expression levels and the number of colonies (both CFU-GM and ln(BFU-E)) that had been normalized by the input CD34+ cell count.

Figure 3.

Correlation of HoxB4 expression levels with CFU-GM and log(BFU-E) normalized for CD34+ cell numbers. CD34+ cells from various samples of mobilized blood from patients with multiple myeloma were stained for the expression of HoxB4 and the cells were processed by EAS for high-resolution immunophenotyping. HoxB4 fluorescence involved a fluorescein-amplified signal. Each sample was also tested for CFU-GM and BFU-E and normalized for the number of CD34+ cells included in the colony forming unit assays. Coefficients of correlation and P values are shown.

Correlation of Expression Levels among Molecules Associated with HRC Function

We also analyzed our dataset to see if the expression levels of the various molecules assessed are associated with each other. By univariate analysis, we found seven clusters of statistically significant (P < 0.05) associations (Table 2 and Fig. 4). There were 14 unique statistically significant correlations among the 120 possible associations tested for the 16 analytes. Multiple linear regression analysis was also performed (Table 3) and indicated that HoxB4, Gab2, and IL-3R significantly predicted E47 levels and that IL-23R and c-Myc predicted GATA-2 expression among other relationships.

Figure 4.

Molecular wiring diagram of HRC. Pathway molecules are shown in brown. Transcriptional regulators are shown in green. Cell surface receptors are shown in red. Colony-forming assays are shown in orange. Direct correlations are represented by blue lines. The single inverse correlations between expression levels is shown with a green line.

Table 2. Univariate correlation analysis
Variable 1Variable 2nrp
Phospho-Akt (ser473)β-Catenin430.4440.003
Phospho-Akt (thr308)GATA-2410.3540.025
CD117 (c-Kit)38−0.3290.044
CD123 (IL-3R)340.4410.009
CD117 (c-Kit)CD130470.3820.009
Table 3. Multiple linear regression analysis
Response variableExplanatory variableParameter estimateStandard errorP
  • a

    Statistically significant

  • b

    Three outliers removed from the analysis



We have used a high-resolution immunophenotyping technology on a flow cytometric platform to assess molecular expression of molecules known to be important in mediating HRC function in CD34+ cells. Because CD34+ HRC constitute a small proportion of the cells in the mobilized blood samples, single cell analysis was needed to obtain these data. As flow cytometry is a relatively insensitive technique for the assessment of low-abundance molecules, enhanced resolution via signal amplification was also required. By combining these analytical modalities, we have been able to make several novel observations.

It should be noted that CD34+ HRC have been previously subdivided into various classes of progenitor/stem cells based on the expression of surface markers (34, 35). Our finding that expression levels of the 16 molecules assessed were unimodal indicate that the differential functional characteristics ascribed to the various subsets within the CD34+ population appear not to be reflected in discrete differences in the expression of these molecules. Nevertheless, a more definitive analysis of this possibility would require simultaneous surface staining that defines the functional subsets and high-resolution immunophenotyping to ascertain the molecular expression levels of molecules important for HRC function in each of the unique CD34+ cellular subsets.

Transplantation for reconstitution of hematopoiesis in myeloablated patients requires the inoculation of cells with the capacity of both self-renewal for long-term effects and early differentiation for limiting the post-transplantation period of vulnerability. These two functions are accomplished by cells, which can be measured by functional assays: CFU-GM and BFU-E for multipotent progenitors that mediate early differentiation and long-term reconstituting assays that measure stem cells with self-renewal capability. The functional assays are expensive, time-consuming, complex, and subjective. For most clinical hematopoietic cell transplant procedures, the only assay currently used to assess the suitability of samples for transplantation is the CD34+ cell count although the CD34 molecule is not involved in either self-renewing proliferation or hematopoietic differentiation.

We found that HoxB4 expression levels are positively correlated with colony forming units, which measures short-term differentiating function. However, this effect was mediated by increased numbers of CD34+ cells. This finding is consistent with previous studies which showed that HoxB4 mediates expansion of HRC in a variety of settings (1, 36–39). Ectopic expression of the molecule in cord blood cells, embryonic stem cells, and CD34+ HRC mediated the expansion of these cells in vitro, while retaining self-renewal capabilities. Our data are consistent with this same effect occurring in myeloma patients who have undergone the mobilization process for autologous transplantation.

Nevertheless, by assessing colony formation with normalization to the input number of CD34+ cells, we have found another effect. Our data suggest that HoxB4 expression is important for the expansion of HRC and at the same time is associated with depreciation in the short-term capacity of the cells to differentiate into the various hematopoietic lineages.

It is important to note that these findings in human mobilized blood samples used for therapeutic transplantation confirm previous experimental findings (40, 41). HoxB4 was overexpressed in purified human cord blood CD34+ cells by retroviral gene transfer (40). The cells with overexpression of HoxB4 had a selective growth advantage when inoculated into nonobese diabetic-severe combined immunodeficient mice but subsequent hematopoietic differentiation of these overexpressing cells was impaired. Similar results were obtained with the ectopic expression of HoxB4 in murine embryonic stem cells and bone marrow cells (41). Thus, our findings translate the insights obtained by experimental manipulations in murine models to therapeutically relevant cells in a clinical setting.

It is important to ascertain that any inoculum used for hematopoietic reconstitution of myeloablated persons contain HRC that mediate both self-renewing proliferation and ready hematopoietic differentiation. We propose that the determination of HoxB4 expression levels in cells used for therapeutic hematopoietic reconstitution can be developed for that purpose. New mobilization protocols might be fruitfully devised that focus on these two types of cells either with their presence in the same sample or in two different samples that could be transplanted together. We propose that the assessment of HoxB4 expression levels in CD34+ HRC be used to ascertain the status of the cells in terms of both self-renewing cellular expansion and short-term hematopoietic differentiation. Nevertheless, it should be clear that in this study we have not defined specific levels of HoxB4 expression that could be used for these purposes. Determination of an appropriate cutpoint will require further investigation.

None of the various associations found among the molecules associated with HRC function have been previously described in HRC. Perhaps the most interesting cluster of associations involves the transcription factor, E47, which is correlated with HoxB4 (another transcription factor), Gab2 (an adapter protein which coordinates signaling downstream of cytokine receptors), and IL-3R (a cytokine receptor involved in HRC function). Using western analysis with cells cultured in vitro, other investigators have shown that Gab2 acts downstream of IL-3R activation (20, 21). Thus, our results elaborate on these initial findings and extend them to samples analyzed ex vivo.

Similarly, the cluster of c-Myc and IL-23R with GATA-2 may be instructive. As GATA-2 expression showed a marginally significant association with CFU-GM but no relationship to HoxB4 or the E47 cluster, the GATA-2 cluster may represent an independent modulator of HRC function. We were surprised that IL-23R correlated with GATA-2 in this study. We had initially included IL-23R as a negative control. We found that IL-23R was expressed on HRC but it had not previously been studied for an association with HRC function. It is possible that IL-23R is involved in HRC function through its interaction with GATA-2.

The strong association of phospho-Akt(ser473) and β-catenin is important because the interaction of the Akt and Wnt signaling pathways had previously been recognized in a study of a tumor line (42). Activation of the Akt pathway has been shown to occur with the differentiation of long-term hematopoietic stem cells into progenitor cells that have the capacity to reconstitute into all lineages but does not possess long-term self-renewal properties (43).

Clearly more samples and more molecules would yield a higher resolution molecular wiring diagram. Nevertheless, our efforts represent the first model to integrate the protein expression levels of these molecules in HRC. Although we have found a variety of correlations, we make no assumptions that any of the associations we have uncovered indicate a causal relationship. Although causality is certainly possible, additional experiments would be required to make this conclusion. Moreover, it should be noted that we have not ascertained the linearity of signal for the detection of the analytes included in this study. Linearity has been established for the technology using a model antigen conjugated to beads (28).

The focus of our study has been the identification of a molecular correlate for HRC function for the purpose of transplantation. However, it should also be emphasized that this same technology can be used to assess the mobilization of HRC. There are now many different regimens that influence the mobilization of HRC. It seems probable that different procedures for mobilization may affect the engraftment capabilities of the HRC. Consequently, it would be optimal to assess mobilization regimen with other molecular analytes in addition to the evaluation of CD34+ cells.


The authors thank Dr. Luis Solchaga and Mr. Robert Fox for help in obtaining patient samples and Dr. Elisabeth Paietta for reviewing the manuscript. Dr. David Kaplan has an equity interest in Pathfinder Biotech, the funder of this study.