Not just a marker: CD34 on human hematopoietic stem/progenitor cells dominates vascular selectin binding along with CD44

Abstract

CD34 is routinely used to identify and isolate human hematopoietic stem/progenitor cells (HSPCs) for use clinically in bone marrow transplantation, but its function on these cells remains elusive. Glycoprotein ligands on HSPCs help guide their migration to specialized microvascular beds in the bone marrow that express vascular selectins (E- and P-selectin). Here, we show that HSPC-enriched fractions from human hematopoietic tissue expressing CD34 (CD34^pos) bound selectins, whereas those lacking CD34 (CD34^neg) did not. An unbiased proteomics screen identified potential glycoprotein ligands on CD34^pos cells revealing CD34 itself as a major vascular selectin ligand. Biochemical and CD34 knockdown analyses highlight a key role for CD34 in the first prerequisite step of cell migration, suggesting that it is not just a marker on these cells. Our results also entice future potential strategies to investigate the glycoforms of CD34 that discriminate normal HSPCs from leukemic cells and to manipulate CD34^neg HSPC-enriched bone marrow or cord blood populations as a source of stem cells for clinical use.

Graphical abstract

Figure 1.

Differential expression of E-selLs on CD34^pos and CD34^neg subsets isolated from the Lin^neg CD38^neg fraction of human BM. (A) Overview of the gating strategy used to isolate CD34^negCD38^neg and CD34^posCD38^neg fractions by fluorescence-activated cell sorting of lineage-depleted (Lin^neg) BM MNCs. Left panel, dot plot represents the cell surface expression of a lineage marker cocktail (including CD2, CD3, CD11b, CD14, CD15, CD16, CD19, CD56, CD123, CD235a, and CD7). Cells residing in the negative fraction (R1) were further gated for CD38-negative cells (R2) (middle panel) and then subdivided into 2 subpopulations based on CD34 expression, CD34^pos and CD34^neg residing in R3 and R4 gates, respectively (right panel). Data shown are representative of n = 4 experiments. (B) Left panel, representative E-Ig staining profile of CD34^pos and CD34^neg subpopulations isolated as depicted in panel A. The shaded curve shows EDTA control (20 mM; on the Lin^negCD38^negCD34^pos subset), whereas dotted red and solid blue curves show the specific binding of CD34^neg and CD34^pos subsets, respectively (n = 4). Middle panel, lysates of CD34^pos BM cells (CD34^pos-BM), Lin^negCD38^negCD34^pos, and Lin^negCD38^negCD34^neg populations isolated from human BM were normalized for total protein level and subjected to western blot analysis. Membranes were blotted with E-Ig, CD44, or β-actin followed by isotype-matched HRP-conjugated mAb for visualization. This is representative of n = 4 independent experiments. Supplemental Figure 1 shows western blots where CD44 was immunoprecipitated from these cell populations and blotted with E-Ig, CD44, and HECA-452. Right panel, flow cytometric analysis of E-selLs expressed on the 2 subpopulations isolated as in panel A is shown as the average percent of expression (above the isotype control) of n = 3 independent experiments. *P < .05 relative to CD34^neg subpopulation. NIH Image J was used to quantify the intensity of western blot bands using the gel analyzer tool; the number displayed represents the density of each band related to the Lin^negCD38^negCD34^neg band as a standard. FSC, forward scatter.

Figure 2.

The sialomucin CD34 is a novel ligand for E-selectin. (A) CD34 was immunoprecipitated from HSPC-enriched lysates of CD34^pos cells isolated from normal UCB cells (CD34^pos-UCB), normal BM cells (CD34^pos-BM), AML BM cells (CD34^pos-AML), or KG1a cells (n = 3 patient samples and KG1a cells). Lysates were subjected to western blot analysis with E-Ig chimera. This is a representative blot of n = 3 experiments. (B) The reciprocal immunoprecipitation assay was performed where E-Ig chimera was used first for immunoprecipitation prior to western blot analysis using CD34 QBend-10 mAb (n = 3). (C-D) AML cells express a unique form of CD34 that does not function as an E-selL. Multiple rounds of E-Ig immunoprecipitations were performed on both normal and AML sample cell lysates, and following the clearance of E-Ig reactive bands, the residual lysates were immunoprecipitated using QBend-10 (CD34-mAb that recognizes all classes of CD34) and subjected to western blot analysis for CD34 (QBend-10) (C) and anti-sLe^x (HECA-452) (D). SN = supernatant. For panel D, as described in the supplemental Materials and methods, only the first elution after E-Ig immunoprecipitation and the CD34 immunoprecipitation are shown. These blots are representative of n = 4 separate experiments. (E) KG1a cells were pretreated with E-Ig chimera or left untreated prior to lipid raft staining with choleratoxin-β (CTB)-AF-594 (green). Fixed cells were then stained with CD34 (Cy5; red) and AlexaFluor-488 streptavidin (blue) toward biotinylated anti-human-Ig to detect E-Ig. The colocalized mask was analyzed using Imaris Coloc software. Cell surface labeling with an isotype control or a secondary antibody alone served as background controls (data not shown). Results are representative of n = 3 independent experiments, 7 fields per experiments; >5 cells per field. Scale bar is 5 μm. (F) CD34^pos cells from AML BM (lower panels) express a CD34 glycoform that does not bind E-selectin, whereas CD34^pos from normal BM (upper panels) counterparts does not. CD34^posCD38^neg lineage-depleted BM cells from either AML or normal BM were prepared for confocal analysis and stained for E-selectin chimera (E-Ig; red), sLe^x expression (HECA-452; green), and CD34 (8G12; blue). Bright field images are also shown. Colocalization analysis was performed using Imaris 7 software to construct the colocalization mask (white). Yellow arrows in the AML images (lower panels) point to spots where CD34 expression is not overlaid with E-selectin binding or sLe^x expression. Results are representative of n = 3 independent experiments. Scale bar is 10 μm. For clarity, panels A-D were performed on CD34^pos samples from normal or AML BM wherever stated and panel F was performed on Lin^negCD38^negCD34^pos cells from normal or AML BM. IP, immunoprecipitation.

Figure 3.

Flow-based binding assays confirm that CD34 is a functional E-selL. (A) Raw data depicting the capturing of E-selectin ligand from KG1a lysate on surface-immobilized antibodies and their subsequent binding to E-selectin by SPR. The lanes of the flow cell represent blank (red line), immobilized CD34-mAb (4H11) (7360 RU; red dashed line), immobilized CD44-mAb (Hermes3) (5200 RU; blue dashed line), and immobilized isotype control (4377 RU; blue line). The mAb immobilization (step 1) is not shown. Lysate injection, arrows mark the start and end of the lysate injection, which is then followed by a buffer washing step. The sensorgrams represent the raw data of the uncorrected RU for the buffer bulk refractive index and nonspecific interactions. Inset, western blot analysis and staining for either CD34 (left panel) or PSGL-1 (right panel) after eluting them from the chip confirming specificity of their capturing by SPR. E-Ig/EDTA injection, the injection involves E-Ig (177 nM) in the presence of EDTA (5 mM) as a control to show specificity of E-Ig to its ligands. E-Ig/Ca²⁺ injection, the injection involves Ca²⁺ (1 mM) to support E-Ig (177 nM) binding to its ligands. (B) Corrected sensorgrams for the buffer refractive index and nonspecific interactions. Data were presented as described in panel A but after subtracting the RU of the buffer refractive index and nonspecific interactions from the standard reference flow cell that contains the isotype control. Furthermore, normalization was applied to even out the difference in the level of mAb-captured CD34 and CD44 as described in the supplemental Materials and methods. k_off is the dissociation rate constant for CD34 and CD44 from their respective mAb, and k_off-apparent is the apparent dissociation rate constant for E-Ig, CD34/E-Ig, or CD44/E-Ig from their respective mAb as well as E-Ig from the complexes CD44·Hermes-3-mAb or CD34·4H11-mAb (n = 3 independent experiments). (C) Immunoprecipitations of CD34 were prepared from lysates of CD34^pos-BM and KG1a cells and spotted on glass slides to test for CHO-E binding using a Stamper-Woodruff assay. Adherent CHO-E cells were counted by light microscopy using an ocular grid under magnification ×20. The data are representative of 1 experiment, and the error bars indicate the standard error of the mean (SEM) of 7 fields per slide on 2 slides for each experiment (n = 3 independent experiments). (D) Blot-rolling assays were performed on western blots of CD34, CD44, CD43, and PSGL-1 immunoprecipitates from KG1a cells stained for HECA-452. CHO-E cells were subsequently perfused over immunoprecipitated glycoproteins at 0.25 dyne/cm². After cell perfusion, the numbers of rolling cells per field were counted (red bars). As a control, CHO-E cells incubated with EDTA (blue bars) or mock-transfected CHO cells (CHO-M) (light green bars) were used. Results shown reflect the average number of rolling cells over the HECA-452 blots of n = 7 membrane preparations from 4 distinct fields of view each. Data are mean ± SEM (error bars). *P < .05; **P < .01; ***P < .001.

Figure 4.

Determination of the dissociation binding constant (K_D) for the binding of E-selLs from HSPC-enriched lysates to E-Ig. The sensorgrams show binding of consecutive injections of E-Ig at 15 μL/min for 240 seconds each at concentrations of 3.9, 7.8, 15.6, 31.3, 62.5, 125, 250, 500, 1000, and 2000 nM that are each spaced by a 60-second buffer washing step (A) over captured E-selLs (CD34, CD44, CD43, and PSGL-1) expressed in KG1a cell lysate (150 mM NaCl) and (B) over captured CD34 from CD34^pos-BM lysate (50 mM NaCl). 563-mAb (10 800 RU for KG1a lysate or 6952 RU for CD34^pos-BM lysate) or MsIgG₁ isotype control (8320 RU for KG1a and 7500 RU for CD34^pos-BM lysate) was immobilized to capture CD34 (left upper panel). Hermes-3-mAb (9300 RU) or MsIgG_2a isotype control (7700 RU) was immobilized to capture CD44 (right upper panel). KPL-1-mAb (11 400 RU) or MsIgG₁ isotype control (8090 RU) was immobilized to capture PSGL-1 (right lower panel). A polyclonal CD43 Ab (15 800 RU) or Goat isotype control (14450 RU) was immobilized to capture CD43 (left lower panel). The sensorgrams presented are corrected for the bulk refractive index and nonspecific interactions using the isotype controls. K_D was determined by fitting the binding isotherm using a steady-state model and the RU_max values just prior to the start of the buffer injection where steady-state conditions were nearly met (inset) and k_off-apparent as described in Figure 3B. Data are representative of n = 3 independent experiments.

Figure 5.

Silencing CD34 leads to a decrease in microvilli and higher rolling velocities with increasing shear stress. (A) Flow cytometric analysis of E-selL expression and E-Ig and HECA-452 binding of scrambled and CD34 siRNA-nucleofected KG1a cells (CD34-KD). This is a representative figure of n = 4 independent experiments depicting the geometric mean fluorescent intensity (G.MFI). (B) Equivalent amounts of scrambled and CD34-KD cell lysates were subjected to western blot analysis and stained for CD34, CD44, CD43, PSGL-1, sLe^x (HECA-452), and E-Ig. Blots are representative of n = 4 independent experiments. (C) Scrambled or CD34-KD KG1a cells were each perfused over CHO-E cell monolayers for 1 minute at 0.28 dyne/cm², and then detachment assays were employed by increasing the shear stress stepwise every 15 seconds. The average number of rolling cells in 4 distinct fields of view for each experiment (n = 4) was counted. (D) Single-cell tracking with Imaris V7.6.4 software was used to calculate KG1a rolling velocity over CHO-E cells from (C) at each shear stress depicted as described. The adhesion histogram is representative of n = 4 independent experiments, and data are reported as the mean ± SEM (error bars). *P < .05. (E) KG1a cells were either transfected with scrambled control siRNA (scrambled control) or with CD34 siRNA (CD34-KD), and 48 hours later, cells were fixed with glutaraldehyde and prepared for TEM analysis as outlined in supplemental Materials and methods. Red arrows point out some of the microvilli-type structures that are evident in the scrambled control but not in the CD34-KD images. These are representative cells of over 30 different cells imaged under each condition.

Figure 6.

E-selectin binds predominantly to O-glycans on CD34. (A) 4H11-mAb (4964 RU) or its isotype control (4036 RU) were immobilized to capture CD34 from KG1a lysates that were treated with neuraminidase (Treated; blue line) or left untreated but subjected to the same buffer-treatment conditions, incubation times, and temperatures (Control; red line). Following CD34 capture, E-Ig was injected at 354 nM. The same surface was used for both the treated and the control binding studies with a surface regeneration step between the 2 runs. The normalized (red dashed line) sensorgram is the same as the control but normalized to the treated sensorgram based on the ratio of accumulated CD34 RU prior to E-Ig injection. (B) For western blot analysis, equivalent amounts of CD34 immunoprecipitates from HSPC-enriched lysates (KG1a or CD34^pos-UCB) were either treated with neuraminidase (+) or not (−) and blotted with either E-Ig (top panel) or CD34 (QBend-10, lower panel). Note that the apparent increase in MW of CD34 is attributed to the loss of the negatively charged sialic acid. (C) SPR analysis of the PNGase F treatment was performed as in panel A using 4H11-mAb (6500 RU) or its isotype control (5810 RU). k_off and k_off-apparent were calculated as described in Figure 3B. (D) CD34 immunoprecipitates were treated with PNGase F and subjected to western blotting for E-Ig (top panel) or CD34 (lower panel) as in panel B. (E) SPR analyses of OSGE treatment were performed as in panels A and C using 4H11-mAb (9000 RU) or its isotype control (6500 RU). (F) Western blot analysis of the treated CD34 immunoprecipitates were performed as in panels B and D. CD34 (Qbend-10) was used as an internal control to confirm N- and O-glycan removal. The sensorgrams presented are corrected for the bulk refractive index and nonspecific interactions using the isotype controls. All results are representative of n = 3 independent experiments.

Figure 7.

A novel CD34 glycoform acts as a P-selectin ligand and its interaction is dependent on O-glycans and tyrosine sulfation. (A) P-Ig was used to immunoprecipitate potential ligands from KG1a and CD34^pos-BM lysates, and the resultant proteins were analyzed by western blot for CD34 (QBend-10 and EP373Y mAb) or PSGL-1 (KPL-1-mAb). Note that CD34 immunoprecipitates were free from any PSGL-1 contamination (n = 3 independent experiments). (B) CD34 and PSGL-1 were immunoprecipitated from KG1a or CD34^pos-BM lysates and then analyzed by western blot for P-Ig binding (n = 3 independent experiments). (C) E-Ig was used to immunoprecipitate E-selLs from the KG1a lysate and resultant proteins were eluted with 30 mM EDTA. The eluate was then immunoprecipitated with CD34 mAbs (clones 4H11 and 581) prior to western blot analysis for CD34, E-Ig, and P-Ig (n = 3 independent experiments). (D) CD34 immunoprecipitates were prepared from CD34^pos-BM and KG1a lysates and spotted on glass slides to test for CHO-P binding using a Stamper-Woodruff assay. Adherent CHO-P cells were counted by light microscopy. The data are a representative experiment, and the error bars indicate the SEM of 7 fields per slide on 2 slides for each experiment (n = 3 independent experiments). (E) Adhesion bar graph representing results obtained for the blot-rolling assay using CHO-P cell rolling (red bars) at 0.25 dyne/cm² (rolling cells per field) over western blots of immunoprecipitated CD34, CD44, CD43, or PSGL-1 from KG1a cell lysates as in Figure 3D. As controls, CHO-E cells were incubated with EDTA (blue bars) or mock-transfected CHO cells (CHO-M; light green bars) were used. The adhesion bar graph is the average of 4 fields of view per experiment from n = 5 independent experiments, and data are reported as the mean ± SEM (error bars). (F) CD34 immunoprecipitates from KG1a lysates were treated with neuraminidase, OSGE, or PNGase F or no treatment followed by western blot for P-Ig binding. Note that CD34 (QBend-10) was used as an internal control (data not shown) as in Figure 6 (n = 3 independent experiments). (G) To inhibit sulfation, KG1a cells were treated with 150 mM sodium chlorate for 72 hours (+; left panel) while KG1a whole-cell lysates were treated with 5 U/mL arylsulfatase for 3 hours (+; right panel) as described in supplemental Materials and methods. Negative controls consisted of the buffers used for the treatment without the sodium chlorate (−; left panel) while KG1a whole-cell lysates were treated arylsulfatase (−; right panel). Following treatment, CD34 was immunoprecipitated from cell lysates, and the resulting proteins were analyzed by western blot for P-Ig binding (top). Note that both treatments abrogated PSGL-1 binding to the KPL-1-mAb from KG1a lysates (middle; KPL-1 is sensitive to the loss of sulfation on PSGL-1), whereas the treatments did not significantly affect CD34 protein levels as indicated by EP373Y-mAb staining (bottom) or PSGL-1 staining (data not shown). Blots are representative of n = 3 independent experiments. (H) Divalent metal ion dependency and binding of P-Ig to PSGL-1 and CD34 captured from KG1a cell lysate at 150 mM NaCl were measured by injecting P-Ig (67 nM) in the presence of 10 mM EDTA followed by consecutive injection of different concentrations of P-Ig (as indicated) in the presence of 2 mM CaCl₂. The experimental conditions are similar to those in Figure 4 with P-Ig injected for 170 seconds interrupted by a 60-second washing buffer step. KPL-1-mAb (11 000 RU) or MsIgG₁ isotype control (5000 RU) were immobilized to capture PSGL-1 (left panel). 563-mAb (6527 RU) or MsIgG₁ isotype control (5000 RU) were immobilized to capture CD34 (right panel) (n = 3). k_off-apparent and K_D were determined as described in Figures 3B and 4, respectively. The sensorgrams presented are corrected for the bulk refractive index and nonspecific interactions using the isotype controls.