Flow cytometric routine CD34 analysis enumerates hematopoietic stem and progenitor cells irrespective of their subpopulations although this might predict engraftment dynamics and immune reconstitution. We established a multi-color CD34 assay containing CD133, CD45RA, CD10, CD38 and CD33. We examined PBSC, donor bone marrow (BMd) and BM of patients 1 year after allografting (BM1y) regarding their CD34 subset composition, which differed significantly amongst those materials: the early CD45RA−CD133+CD38low subpopulations were significantly more frequent in PBSC than in BMd, and very low in BM1y. Vice versa, clearly more committed CD34 stages prevailed in BM, particularly in BM1y where the proportion of multi-lymphoid and CD38++ B-lymphoid precursors was highest (mean 59%). CD33 was expressed at different intensity on CD45RA±CD133± subsets allowing discrimination of earlier from more committed myeloid precursors. Compared with conventional CD34+ cell enumeration, the presented multi-color phenotyping is a qualitative approach defining different CD34 subtypes in any CD34 source. Its potential impact to predict engraftment kinetics and immune reconstitution has to be evaluated in future studies.
Hematopoietic stem and progenitor cells (HSPC) were originally described by simple flow cytometric methods using only CD34 to characterize them.1, 2 Since 1996 the two-color ISHAGE protocol focusing on CD45 and CD34 has served as the standard for CD34+ cell enumeration.3 Addition of 7-AAD to define cell viability, and the use of beads for standardized quantification expanded the test to a single-platform three-color analysis tool to enumerate CD34+ cell numbers in bone marrow (BM), blood or cord blood.4, 5 In the clinical setting, these analyses have not changed and address only CD34+ cells as such. However, one important issue concerning repopulation and engraftment after hematopoietic stem cell transplantation is to define distinct subsets of CD34+ cells.6, 7, 8 Markers like CD45RA and CD38 were suggested in research experiments.9, 10, 11, 12 Expression studies of CD34 with CD10, CD38, CD45RA or CD90 led to the discovery of functional HSPC-subpopulations and to the presentation of the classic model of the human hematopoietic lineage tree.13, 14 Some studies correlated the co-expression of CD38 and CD71 with neutrophil and platelet recovery.6, 15 CD133/Prominin-1 was included only recently into a panel of markers to define different CD34 subsets.16 Based on these studies, Görgens et al.17 suggested a revision of the most recent model of human hematopoiesis, the composite model (Figure 1), originally proposed by the Jacobsen group.14, 18 According to this, multi-potent progenitors (MPP) represent early CD34 developmental stages enriched within the CD45RA−CD133+CD38lowCD10− cell fraction most of which dividing asymmetrically.19 Tracking CD133 segregation in functional single cell analysis, the authors showed that MPP create pairs of daughter cells, the lymphoid-primed multi-potent progenitors (LMPP) and the erythro-myeloid progenitors (EMP). Whereas LMPP upregulate CD45RA and remain CD133+ (CD45RA+CD133+CD38lowCD10−), EMP remain CD45RA− but lose CD133 expression (CD45RA−CD133−/lowCD38+CD10−). LMPP harbor the potential to create neutrophil and macrophage progenitors (GMP) as well as CD10 upregulating multi-lymphoid precursors (MLP). Among others, these MLP create CD34+CD19+ B-lymphoid progenitors (BLP), which downregulated CD133 (CD45RA+CD133−CD38++CD10+). Based on the definition of these subsets, we established a multi-color approach to define and enumerate CD34 subpopulations in the three different HSPC sources. CD34, CD45, 7-AAD and scatter properties were used to define viable CD34+ cells, which were then further classified using CD45RA, AC133, CD7, CD10, CD19, CD3, CD33 and CD38.
Materials and methods
All cell specimens used in this study were obtained for routine CD34 enumeration. Donor BM samples (BMd, n=31) were from healthy allogeneic BM donors aged from 2 to 48 (median 26) years. Patient BM specimens (BM1y, n=21) were from biopsies routinely drawn for clinical examination 1 year after allografting. Six of these were paired samples, that is, we analyzed both BMd and BM1y from these patients. The BM recipients (11 males, 10 females) had been diagnosed with AML (4), ALL (6), MDS (4), CML (2), RCC (2), SCID (1), SAA (1) and Hyper IG E syndrome (1). Their median age was 10 years (range 0.8–18). Conditioning regimens were myeloablative (n=11) and reduced intensity (n=10). Fifteen BM recipients had received antithymocyte globulin. PBSC samples (n=35) were from autologous stem cell collections from adult patients (22 males, 13 females) with multiple myeloma (n=19) and Non-Hodgkin's lymphoma (n=16), with a median age of 54 years (range 22–73). The mobilization regimen comprised Cht (chemotherapy) and hGF (hematopoietic growth factors) n=13), Cht+hGF and plerixafor (n=6), hGF alone (n=5) or hGF and plerixafor (n=11). Informed consent was obtained from all patients and the extended staining experiments had been approved by the local ethical committee.
Cell processing and immune staining
Blood cell counts were obtained from a Sysmex KX-21N (Sysmex Corporation, Kobe, Japan). If necessary, cells were diluted with Dulbecco's PBS (Carlsbad, CA, USA) and the WBC adjusted to 5–15 × 106 cells/mL prior to immune staining. All monoclonal antibodies were used at pretested concentrations and after respective compensation. Isotype controls and fluorescence-minus-one analyses were used to define gating and compensation.
Single platform protocol
In total, 100 μL of the cell sample were reverse-pipetted into a Trucount tube (BD Biosciences, San Jose, CA, USA). After adding the monoclonal antibodies cocktail, cells were mixed and incubated light-shielded at room temperature for 20 min. RBCs were lysed by adding 2 ml of ammonium chloride working solution (BD Biosciences) for 10 min before samples were analyzed on the flow cytometer. The monoclonal antibody cocktail contained the stem cell enumeration kit (BD Biosciences) comprising CD45 FITC, CD34 PE and 7-AAD, as well as the following monoclonal antibody: AC133-1 APC, CD7 PeCF594, CD10 BV421, CD19 APC-Cy7, CD38 PE-Cy7, CD45RA BV510, CD3 PerCPeFl710, and CD33 APC-R700. For detailed information please see Supplementary Table 1.
Acquisition of 150 000 CD45+events was done on a FACS Fortessa (BD Biosciences) equipped with four solid state lasers with excitation wave lengths (nm) of 488, 405, 561 and 640. The FACSDiVa 6 software (BD Biosciences) was used for cell acquisition and data evaluation. For quality control of the instrument’s performance, CS&T beads (BD Biosciences) were used at least weekly.
Viable WBC were defined by their CD45 expression, negativity for 7-AAD, and typical position in the forward- and side scatter (FSC/SSC) dot plot. According to the ISHAGE guidelines,4 viable and true HSPC were determined by their positivity for CD34, their weak expression of CD45, their typical position in the lympho-monocytic area of the FSC/SSC dot plot and their negativity for 7-AAD. To define subpopulations, CD34+ cells were first divided into an earlier CD45RA− and a more committed CD45RA+ cell fraction. These were then separately depicted in a CD133 vs CD10 contour plot (Figure 2, Supplementary Figure 1D). The resulting subpopulations were examined for their expression of CD38, CD33, CD10 and CD7. Beads were double-gated in two different dot plots (APC vs SSC and APC-Cy7 vs FITC) to exclude false-positive events, and the following formula was used to calculate the number of target cells/μL:
To calculate absolute cell numbers in donor BM, an additional dilution factor of 1.1 was considered to compensate for the anticoagulant added.
Differences in cell counts (both absolute and relative) between the three HSPC sources were assessed with the unpaired one-sided Wilcoxon signed-rank test. A P-value <0.05 was considered statistically significant (*<0.05, **<0.01, ***<0.001). Mean values (±s.d.) are provided in the text. In depicted Box-and-Whisker plots, boxes range from first to third quartile (containing 50% of data points). The median value is indicated by a thick horizontal line. Whiskers extend to the most extreme data point, which is <1.5 times the interquartile range (=box height) away from the box, and indicate the range that contains 95% of data points in a normally distributed sample. R version 3.2.0 (2 015-04-16) was used for all statistical analyses.
Enumeration of total HSPC irrespective of CD34 subtypes revealed significant differences between the three cell sources (Figure 3a). Results were expressed as absolute numbers (viable CD34/μL) and as relative values (viable CD34+ cells as percentage of viable WBC). The mean percentage (±s.d.) of total CD34+ cells was highest in BMd (2.1%±2), followed by BM1y (1.1%±0.6) and by autologous PBSC (0.8%±0.9). In terms of absolute CD34 numbers, the highest values were obtained in PBSC (1402/μL±1049), followed by BMd (794/μL±753) and by BM1y (255/μL±253). The majority of PBSC samples described in the present work was from adult patients diagnosed with different lymphoid malignancies. Since most of the pediatric patients had been allografted with BMd, only few PBSC samples had been obtained and analyzed from healthy mobilized donors, but their comparison revealed a similar distribution of CD34 subtypes in patient and donor PBSC (data not shown).
CD34 subtyping was always started with CD45RA, as this marker allows definition of two distinct subgroups in all materials, separating earlier CD45RA− from more committed CD45RA+ HSPC (Figure 2, Supplementary Figure 1). Both subpopulations were then further evaluated in CD133/CD10 contour plots where they formed at least six distinct CD34 subpopulations. Between the three cell sources analyzed, we observed considerable differences regarding the composition of CD34 subsets, both for absolute cell numbers and relative values (Figure 3b). The mean cell proportion (in % CD34+ cells±s.d.) of the early MPP cells was significantly higher in PBSC (42%±13.7) than in BMd (16%±8; P<0.001). In BM1y, their frequency was only 2.5%±1.8, which was significantly lower than in BMd (P<0.001) and roughly 17-fold less than in PBSC. Analogous results were obtained for absolute MPP numbers, which were significantly higher in PBSC (608/μL;±475) than in BMd (138/μL±156; P<0.001), and in BMd than in BM1y (5/μL±6; P<0.001).
MPP differentiate to either LMPP (CD45RA+CD133+) or EMP (CD45RA−CD133−/low). Both subsets showed a higher CD38 expression than MPP supporting their higher differentiation (Figure 4). As shown in Figure 3b, the mean frequency of LMPP was similar between PBSC and BMd (25.6%±11.1, and 23.7%±10.1; ns), but significantly lower in BM1y (16.3%±8.3) than in BMd (P<0.01). Absolute LMPP cell numbers differed significantly between PBSC (383/μL±412) and BMd (170/μL±158; P<0.01), and between BMd and BM1y (33/μL±28; P<0.001). This may suggest a higher proportion of neutrophil progenitors in favor of BLP in the LMPP subfraction of PBSC compared with BMd. EMP that give rise to erythrocytes, megakaryocytes and granulocytes other than neutrophils, showed results comparable to those obtained for LMPP. Their mean frequency was similar in PBSC (24.5%±8.7) and BMd (19.5%±6.5; P<0.05) but differed significantly between BMd and BM1y (11.9%±7.3; P<0.001). Absolute EMP numbers differed clearly and were 331/μL±261 for PBSC vs 162/μL±176 for BMd (P<0.001), and 24/μL±25 for BM1y vs BMd (P<0.001). In terms of LMPP and EMP, BM1y thus mediates an impression of exhaustion when compared with BMd. Late myeloid progenitors (late GMP) with a CD45RA+CD133−CD33+CD10− phenotype differed mainly with regard to relative values, which were significantly lower in PBSC (4%±3.1) than in BMd (8.5%±4.1; P<0.001), but similar between BMd and BM1y (11.1%±6.5; ns). In terms of absolute values, they were 51/μL±62 in PBSC vs 72/μL±91 in BMd (ns), and slightly lower in BM1y (23/μL±20) than in BMd (P<0.01). Owing to the low frequency of the CD133+ MLP, this CD34 subset was evaluated together with the CD133− BLP. These cells were hardly detectable in PBSC (5%±7.9) but clearly present in BMd (31.9%±15.1; P<0.001). In BM1y, they represented the largest CD34 subfraction (58.5%±17.6), which was significantly higher than in BMd (P<0.001).
Despite the rather low proportions of the CD133dim MLP in the BM samples, it has to be noted that this CD34 subset, when expressed as percentage of the MLP/BLP fraction, represented a clearly lower median cell proportion (P=1.66e-09; one-sided unpaired Wilcoxon rank sum test) in BM1y (3.6%) than in BMd (11.9%). A representative example is depicted in Figure 2. In terms of absolute CD10+ stem cell numbers, significant differences were observed between PBSC (43/μL±83) and BMd (255/μL±302; P<0.001), whereas the values were similar between BMd and BM1y (171/μL±191; ns). Out of the 31 BMd and 21 BM1y specimens analyzed, six were paired samples, that is, we examined BMd and BM1y pairs from the same patients. The results were virtually identical to those obtained from the whole groups: The median proportions of the CD34 subsets for BMd/BM1y were 12.1%/1.8% (MPP), 22%/17.4% (LMPP), 20.1%/10% (EMP), 5.5%/9.1% (late GMP) and 34%/65.2% (MLP and BLP).
The expression intensity of distinct markers often correlates with differentiation as shown for CD38 (see below). Such differences were also observed for CD33 and CD133. Expression of CD133 on MPP and LMPP was generally higher in PBSC than in BMd and BM1y (not depicted), suggesting that these cell stages are more differentiated in BM than in PBSC. Nevertheless, it was possible to distinguish the different CD133+/− subpopulations in all cell samples (Figure 2). The myeloid marker CD33 was expressed in all CD34 subfractions, although it was weaker on MPP in PBSC with high CD133 expression than on MPP in BM with weaker CD133 expression (not depicted). Distinct CD10+ (and CD19+) HSPC subsets were detectable among both the CD45RA+ and the CD133− progenitors, and the CD33 expression was clearly higher in BM1y than PBSC (Figure 5). In all cell sources, a potential co-expression of CD7 as a marker of T- and NK-cell progenitors was only seen on 0–2.8% of CD34+ cells (see Supplementary Figure 1 D3). As non-specific staining could not be excluded, this subtype was not pursued any further.
We used the PE-Cy7-labeled H7 clone of CD38 in all experiments performed. Virtually, all CD34+ cells were positive for this Ab, albeit at clearly different intensity (Figure 4). MPP showed the lowest expression intensity, followed by LMPP and EMP. The intensity was higher among CD45RA+ late myeloid precursors, and highest among the CD45RA+CD10+CD19+ BLP. This differential CD38 expression was generally observed in all specimens examined although the differences were not always as clear as depicted in the BM1y sample shown in Figure 4 (see also Supplementary Figure 2). Owing to the considerable overlap between the different CD34 subgroups, CD38 was never used as first-line Ab for subgroup discrimination.
Based on previously presented data,13, 14, 17, 18, 19, 20, 21, 22 we extended our recently described five-color CD34 enumeration protocol23 to an 11-color single-platform analysis tool, which we used to describe at least six distinct CD34 subpopulations differing in terms of maturation und lineage commitment. We have thus reenacted and translated into clinical routine application what the above authors described in their extensive functional analyses, and we show that the composition of the CD34 subsets differs substantially between the three CD34 sources examined. We assume that routine application of the presented protocol will allow high from medium or poor quality transplant materials to be distinguished and may thus predict engraftment characteristics as addressed by several authors,6, 7, 8, 15, 24 but this will have to be proven in future studies. The present results show clearly that PBSC contain significantly higher proportions of MPP, LMPP and EMP than BMd. This may be the reason for the more rapid sustained engraftment after infusion of PBSC instead of BM.25, 26 In BM1y, these earlier CD34 subtypes are again significantly lower than in BMd. In contrast, the more committed CD34 stages, that is, MLP and particularly BLP, are significantly higher in BM1y than in BMd, whereas they are hardly present in PBSC. Our data may be biased by the fact that the PBSC were from autologous patients with a median age of 54 years, whereas the BMd specimens were from younger and healthy donors (median age 26 years). However, one can assume that PBSC from younger and healthy donors (or from cord blood samples) display a more immature rather than a more committed phenotype thus making the difference between PBSC and BM early CD34 subtypes rather larger than smaller.
In normal BM, B-cell precursors may represent a considerable CD34 subfraction, which in our BMd cohort, accounted for median 32%. These cells represent predominantly (~90%) the more mature CD133−CD10+CD19+ BLP, but only few MLP still exhibiting a weak CD133 expression. During mobilization, such rather committed precursors hardly appear in the circulation, which is in contrast to MPP and EMP of which high proportions enter the blood stream. Reasons for the differential mobilization of the various CD34 subsets from the BM into the circulation may be that the earlier CD34 subsets show a better stimulation response to the myeloid growth factor G-CSF than late GMP and B-lineage progenitors. However, it can also not be excluded that the earlier progenitors are less adherent to the BM niche because they prevail not only in mobilized PBSC but also in non-mobilized blood or in cord blood.18, 27
The differences described between PBSC and BMd in terms of CD34 subset composition confirm an earlier observation of higher proportions of CD45RA− progenitors in PBSC than in BM. Cell sorting and culture in semisolid medium revealed that the CD45RA− vs CD45RA+ cells formed compact vs dispersed colonies originating from earlier and more committed myeloid progenitors, respectively.10 A differential mobilization is also supported by the observation that mobilized PBSC are not or hardly in G2S phase,28, 29 which is in contrast to BMd CD34+ cells and, particularly, to BM1y cells.30 In 1997, Anderlini & Körbling assumed a more primitive phenotype of CD34+ cells contained in PBSC compared with BM.25 Furthermore, they postulated a faster engraftment after PBSC transplantation compared with BM allografting as described before.26 This was later confirmed by different authors, who reported no differences in both overall and event-free survival between PBSC and BM, but observed a clearly faster neutrophil and platelet engraftment after infusion of PBSC.31, 32, 33 The most likely reason for this is the higher number of LMPP and EMP giving rise to neutrophil- and platelet-forming cells, respectively.17 It is also noteworthy that the GMP which form monocytes and neutrophils, are contained in the LMPP cohort, which, in BM, contains significantly more B-lineage progenitors. The relative fraction of neutrophil progenitors should therefore be clearly larger in PBSC than in BM although similar frequencies of LMPP (which can give rise to both neutrophil and B-lineage precursors) were found in these two stem cell sources.
CD38 was used by several authors to distinguish earlier from more committed CD34 cell stages.24, 34, 35 Despite a relatively high expression on all HSPC subsets, its mean fluorescence intensity values were clearly lower on CD133+ HSPC subtypes than on more committed CD34 stages. However, owing to the substantial overlap between the different CD34 subtypes, CD38 could never be used to unambiguously separate early from more committed progenitors. It was therefore important that CD34 subtyping was started with CD45RA and followed by CD133 and CD10. When the resulting subsets were then analyzed for their CD38 mean fluorescence intensity, the gradational increase became obvious, ranging from the lowest expression among the young MPP, over LMPP and EMP to the highest among BLP.
One year after allogeneic BM transplantation, MPP accounted for only 2.5% of CD34+ cells in BM1y, which is ~6 times and 17 times lower than in BMd and PBSC, respectively. In contrast, MLP and BLP in particular represented almost 2/3 of CD34+ cells in BM1y, which is significantly more than in BMd. One possible explanation for the marked relative and absolute loss of MPP may be the ‘replicative stress’. Another reason could be the destructive effect of conditioning radio- and/or chemotherapy,36 leading to an unfavorable impact on the homing capabilities and thus to a disturbance of asymmetrical cell division, which was reported for MPP, but not for the more committed CD34 subfractions.19, 29 Both observations, the loss of earlier and the gain of more committed CD34 subtypes are in line with other reports. Selleri et al.37 found an eightfold decrease of long-term culture initiating cells in the BM of patients as compared with donors, and Thornley et al.30 described a higher telomere loss in the BM of patients with full donor chimerism. Woolthuis et al.38 reported a loss of quiescence and impaired function of CD34+/CD38low cells 1 year after autologous stem cell reinfusion. These reports are in line with our observation of a significant increase of CD133dim MLP and CD133− BLP in BM1y, which was also described by Wolf et al.39
BM biopsies 1 year after transplantation are usually only performed after allogeneic, but not after autologous transplantation. However, Bhatia et al.40 examined BM of Non-Hodgkin and Hodgkin lymphoma patients 0.5–3 years after autologous PBSC reinfusion. They also reported a profound and persistent reduction in primitive and an increase of more committed progenitors. This suggests that our observed high proportions of BLP are not a result of allogeneic B-cell activation but rather the consequence of profound B-cell depletion in the context of high-dose chemotherapy. This is supported by the fact that all our patients displaying high numbers of BLP were free of chronic GvHD, in contrast to one patient with chronic GvHD who showed very low MLP and BLP counts.41 As nearly all our patients who underwent allogeneic transplantation during the past 2 years had received BMd but not PBSC, we cannot establish whether the MPP compartment in BM1y would be significantly larger after transplantation of allogeneic PBSC instead of BMd. However, the data presented by Bhatia et al.40 and own preliminary results from two patients who had received allogeneic PBSC comprising MPP numbers clearly higher than those analyzed in BMd suggest that the proportion of early precursors is strongly reduced even after transplantation of allogeneic PBSC.
We conclude that the presented analysis can identify and enumerate distinct CD34 subfractions in any conventional CD34 cell source. This approach may provide a solid basis for future studies to determine the impact of different CD34 subsets in the graft as well as in post-transplant BM on engraftment kinetics and immune reconstitution. Whether or not the analysis will allow prediction of engraftment kinetics on a routine basis remains to be examined.
We would like to thank Daniela Scharner, Dijana Trbojevic and Elke Zipperer for their excellent commitment and input when establishing the described flow cytometric assay, and for their continuous support with cell preparation and data acquisition and evaluation. Dieter Printz is particularly acknowledged for his technical support in all issues of flow cytometry.
About this article
Supplementary Information accompanies this paper on Bone Marrow Transplantation website (http://www.nature.com/bmt)