Rare cell analysis in flow cytometry has always been a challenge due to the many obstacles and artifacts, which accompany the technique (1–3). For this reason careful experimental setup and controls have to be performed to assure data integrity. Nevertheless, when it comes to analysis of low frequency cell populations, flow cytometry reaches certain limits. In our case it is the incomplete doublet exclusion by plotting of forward and sideward scatters areas, heights, and widths (FSC- and SSC-A/-H/-W).
Stem cell-like memory T cells are suggested to crucially contribute to long-term T-cell memory. They have been first discovered in a graft-versus-host disease mouse model and are reported to express high levels of stem cell antigen-1 (Sca-1) (4). Looking for human stem cell-like memory T cells, we analyzed bone marrow samples for the expression of stem cell markers on T cells. Our staining procedure comprises Fc blocking and our analysis includes dead cell, CD14/CD16/CD19, and doublet exclusion. Doublets exhibit a higher signal width or area to height ratio compared to single cells (singlets). There are two commonly used options to gate for singlets using forward and sideward scatters. The first is to plot FSC-A vs. FSC-H. Events deviating from the diagonal are doublets. The second option is to perform a sequential gating. First FSC-H is plotted vs. FSC-W and then SSC-H vs. SSC-W. In both dotplots, events with a low signal width are to be gated in order to obtain singlets. Using this staining and gating procedure we observed CD34+ T cells at low frequencies (Fig. 1a), between 0.1% and 0.6% of T cells. The addition of 2 mM EDTA to the staining buffer did not alter the frequency of CD34+ T cells (data not shown). These CD34+ T cells also showed a higher expression of other stem cell markers such as CD133, c-Kit, and CD59 than other T cells (Fig. 1b). Comparing paired bone marrow (BM) and peripheral blood (PB), we observed that CD34+ T cells are a BM phenomenon, as they were almost undetectable in PB (Fig. 1c).
However, when we tried to sort human BM-derived CD34+ T cells using fluorescence-activated cell sorting (FACS), the purity control afterwards demonstrated an unexpected result (Figs. 2a and 2b). The sorted CD34+ T cells were either CD3+ or CD34+, but hardly double positive. Here the proportion of CD3+ to CD34+ cells was in the range of 60:40. This indicates that BM-derived CD34+ T cells are just doublets of T cells sticking to stem cells. Probably, CD3+ CD34+ events sorted during FACS split off due to the shear stress of the procedure and were then measured as single cells.
Next, we utilized DNA content analysis with 4′,6-diamidino-2-phenylindole (DAPI). This method allows the identification of cell doublets based on the DNA-binding dye signal (5). The principle is the same as for the forward and sideward scatters. Thus plotting DAPI-A vs. DAPI-H and gating on the diagonal should result in single cell events. But again most CD34+ CD3+ events, which have already passed the forward and sideward scatters based singlet gates, did not show a DAPI-A to DAPI-H ratio indicative of doublets (Fig. 2c). However, most CD34+ CD3+ events exhibited a DNA content of 4N or more indicating their doublet character. When we gated on cells with a DNA content between 2N and 3.8N, almost no CD34+ CD3+ events could be detected, as shown in Figure 2d. By doing so all resting and most proliferating cells are incorporated into analysis, whereas polyploid events (4N and more) are excluded. Hence, here the nature of the DAPI signal itself does not help to further exclude doublets, but the DNA content information is already a good hint.
For further validation we applied microscopy-based high content screening. This technique allows automated acquisition and analysis of fluorescent labeled cells. We visualized 503,641 living BM cells, shown by Calcein AM violet fluorescence. About 78,006 of these were T cells (CD3+) and 2241 were stem cells (CD34+). But not a single optical validated cell was positive for CD3 and CD34. However, sometimes T cells were found in close contact to stem cells, most likely accounting for the doublets found in flow cytometry (Fig. 2d).
To see whether this doublet issue can be observed for other cells as well we analyzed BM samples by flow cytometry for the co-expression of CD19, the common B-cell marker, on T cells. In a similar manner, we observed CD3+ CD19+ events at low frequencies (Supporting Information Figure), indicating that this is a cell marker-independent problem. Hence, this experiment demonstrates that this kind of artifact holds true, not only for CD34+ CD3+ events but for other marker combinations as well.
Now, why can these doublets pass both doublet exclusion strategies, based on FSC/SSC and DAPI? The remaining doublets seem to mimic singlets in terms of signal widths/duration. This may be the case, when doublets are oriented in line or coplanar with the laser beam. Then the time to pass the laser would be the same as for singlets, so that signal width and area to height ratio are not increased. However, DNA content analysis can at least reveal the polyploidy of these rare events.
Altogether we conclude that CD34+ T cells are not present in human bone marrow. The observation we made with flow cytometry was caused by the incomplete doublet exclusion by forward and sideward scatters. Still most doublets can be detected and excluded from analysis with this approach, but a small number of doublets remains and may disturb rare cell analysis.
To analyze whether the problem of incomplete doublet exclusion would occur using another type of flow cytometer, we measured bone marrow mononuclear cells with a Beckman Coulter Navios flow cytometer. In fact, we observed CD34+ CD3+ events despite regular doublet exclusion in a similar manner as with BD LSR devices (data not shown). Thus, one may speculate that the problem of incomplete doublet exclusion is a general problem of flow cytometers using hydrodynamic focusing.
So far no article claims the existence of CD34+ T cells in healthy subjects, but we are aware that other groups were investigating in that direction. With this note we want to inform the community about our pitfall, so that others do not commit the same mistake and lose time, effort, and money on this. Of course, we cannot eliminate the possibility that CD34+ T cells are present in other species, tissues, or experimental setups such as after activation. In this report, we clearly demonstrate that CD34+ CD3+ events are an artifact and are as prevalent as CD19+ CD3+ events.
Taken together, the described phenomenon of false positive events due to doublets is an important issue in rare cell analysis. Especially when a rare subset of cells appears to express a marker, which is typical for another cell type in the same sample, extra caution is necessary. One possible approach to lower the risk of false positive events due to doublets is a 5-color-10-antibodies protocol employing an elaborate combination of positive and negative gatings (6). In case of doubt optical validation is a trustful tool. Microscopy-based high content screening is one appropriate way to do this. The technique allows fluorescent image acquisition of high cell numbers to warrant a reliable statistical power. And usually people tend to trust things they can see.