CD8α⁺ mouse spleen dendritic cells do not originate from the CD8α^- dendritic cell subset

The CD8α⁺ and CD8α^- subsets of dendritic cells (DCs) from mouse spleen are of special interest because, as freshly isolated cells, they differ in functional characteristics.¹  However, the developmental origin of these DC subtypes and their relationship to each other has been controversial. The original view that splenic CD8α⁺ DCs were of lymphoid origin and that splenic CD8α^- DCs were of myeloid origin has been shown to be incorrect.^2,3  Differences in the cytokine and transcription factor requirements suggest that the developmental pathways involved differ.^4-8  Culture of purified splenic CD8α⁺ DCs or splenic CD8α^- DCs has failed to show any transformation from one to the other,⁹  though evidence indicates that CD8α can be induced to at least moderate levels on some DC subtypes.^10,11  An in vivo kinetic study, using continuous 5-bromodeoxyuridine (BrdU) labeling, showed no sign of a precursor-product relationship between CD8α^- and CD8α⁺ DCs.⁹ 

In direct conflict with this is the conclusion in a recent paper by Martinez del Hoyo et al.¹²  They isolated CD8α^- DCs, transferred them intravenously into a Ly5-disparate, nonirradiated recipient, and found 3 to 4 days later CD8α⁺ DCs of donor origin. They concluded that CD8α⁺ DCs normally originate from CD8α^- DCs—the 2 are merely maturation stages of the same lineage. In view of the conflict, we have now carried out similar experiments. Although we did obtain some CD8α⁺ DCs on transfer of a CD8α^- DC preparation, recoveries were extremely low, and we argue this is an uncommon event that might be attributed to a contaminant precursor. We find that most of the immediate precursors of CD8α⁺ DCs are cells lacking normal DC characteristics.

C57BL/6J WEHI (Ly5.2) mice 6 to 8 weeks of age were used as donors, and C57BL/6 Pep^3b mice (Ly5.1) of a similar age range were used as nonirradiated recipients.

The method used, described previously,¹³  was similar to that of Martinez del Hoyo et al,¹²  except that fluorescence-activated cell sorting (FACS) rather than immunomagnetic microbead selection was used in the final separation. Samples of the original spleen suspension (after red cell and dead cell removal¹⁴ ) and of cells at all stages of the procedure were retained for counting and assay of precursor activity. Briefly, spleens were digested for 20 minutes at room temperature with collagenase-DNAase then were treated with ethylenediaminetetraacetic acid (EDTA). All subsequent procedures were performed at 0°C to 4°C in an EDTA-salt solution. Light-density cells, 5% of the total nucleated spleen cells, were selected by centrifugation in a 1.077 g/cm³ Nycodenz medium. Non-DC and plasmacytoid cells were depleted by incubation with rat monoclonal antibody (mAb) to CD3, CD45R (B220), Thy-1, Gr-1, and erythrocytes (TER-119), and they were removed using antirat immunoglobulin immunomagnetic beads. Autofluorescent cells were removed by rapid presorting of the unlabeled preparation or by gating out at the time of sorting.¹³  Control experiments indicated an average 50%, and never more than 70%, of DCs were lost during these enrichment procedures. The resultant DC-enriched preparation, approximately 2% of the original spleen, was usually stained for CD11c (N418-fluorescein), major histocompatibility complex 2 (MHC 2) (M5/114-Alexa 594), and CD8α (YTS169.4-Cy5) and for the CD11c^hiMHC 2^hiCD8α^- fraction selected. Alternatively, the DC-enriched preparation, 50% to 60% pure and representing approximately 2% of the original spleen, was usually stained for one of these markers, and cells with negative, medium, and high fluorescence above background were sorted. Reanalysis of the CD11c^hiMHC 2^hiCD8α^- fraction gave a purity of 99%, with no CD8α⁺ DCs detected.

The assay used was based on that used by Martinez del Hoyo et al.¹²  Samples of various DC-enrichment fractions were transferred intravenously (1-10 × 10⁶ cells per mouse) into 2 or more nonirradiated Ly5.1 recipients. Three days later, the spleens of recipient mice were pooled. DCs were then extracted and enriched by density separation and immunomagnetic bead depletion, as described,¹³  to allow clean analysis of donor-derived DCs. Cells were counted and stained for donor-type Ly5.2 (ALI-4A2-biotin with streptavidin-phycoerythrin or streptavidin-Alexa 594 second stage), for CD11c (N418-fluorescein), for CD8α (YTS169.4-Cy5), and in some experiments for CD205 (NLDC-145). The number of donor-derived CD8α⁺ DCs produced per cell transferred was then determined as a measure of precursor activity.

We first attempted to repeat the experiment whereby the transfer of purified CD8α^- splenic DCs produced CD8α⁺ DCs.¹²  However, to obtain high purity and to segregate different levels of marker expression, we used FACS rather than immunomagnetic particle separation in the final separation of CD8α^- DCs. We additionally selected cells bearing relatively high levels of MHC 2, therefore defining more tightly the DC population. Purity was close to 99% on reanalysis, so approximately 1% of the preparation represented possible contaminants. These cells (3-5 × 10⁶) were then transferred intravenously into nonirradiated Ly5.1 recipients. A DC-enriched preparation was made 3.5 days later from the pooled recipient spleens, and the donor-derived (Ly5.2⁺) cells were gated and analyzed.

In 3 of 7 transfer experiments, donor-derived DCs were detected in the recipient spleens. From 5% to 30% of these donor-derived DCs were CD8α⁺ (Figure 1) and DEC-205⁺, whereas the rest remained CD8α^- or stained at only low levels for CD8α. Similar results were obtained at days 2 and 4. Thus, we sometimes did see the generation of CD8α⁺ DCs from a CD8α^- DC fraction, though, in contrast to the previous study,¹²  only a minority showed the acquisition of CD8. Given that most splenic DCs turn over in 3 days,⁹  most donor-derived DCs should have been CD8α⁺ if they originated from the CD8α^- DC subset.

In addition to the low proportion of DCs that became CD8α⁺, a disturbing aspect of the experiment was the low overall recovery rate of the transferred DCs in the spleens of the recipient mice (Table 1). This was true at all time points studied. The highest recovery rate measured in these experiments was 0.33% of the CD8α^- DCs transferred, of which only approximately one quarter were CD8α⁺ DCs (Table 1). Even allowing for a 50% loss of DCs in the enrichment procedure preceding analysis, at most 0.16% of the transferred DCs produced CD8α⁺ DCs in the recipient spleens.

This is in line with the reported values of Martinez del Hoyo et al,¹²  from which it can be calculated that only 0.3% of the transferred cells were recovered as CD8α⁺ DCs. Because this is within the contaminant level of the transferred CD8α^- DCs, it was not proven that it was the CD8α^- DCs themselves that produced the CD8α⁺ DCs.

It could still be argued that this observed production of CD8α⁺ DCs, though derived from only 0.3% of the CD8α^- DC fraction injected, was representative of the other 99.7% transferred but lost. If all the immediate precursors of CD8α⁺ DCs were indeed CD8α^- DCs, such precursor activity, as measured by the ability to produce CD8α⁺ DCs in the spleen 3 days after transfer, should separate with CD8α^- DCs and therefore be more than 100-fold enriched in a pure CD8α^- DC fraction. The distribution of such immediate CD8α⁺ DC precursor activity in fractions sampled during the steps of DC isolation was therefore determined after intravenous transfer of a counted number of cells. Precursor activity was determined as specific activity (donor-derived CD8α⁺ DCs per recipient spleen per 10⁶ donor cells transferred) and then calculated as total activity (specific activity × 10^-6 multiplied by the number of cells in each separation fraction from one donor spleen) (Table 2). Specific activity allows a measure of precursor enrichment, and total activity allows a measure of precursor recovery.

To gauge the enrichment and recovery of CD8α⁺ DC precursors concomitant with CD8α^- DC enrichment, the total CD8α⁺ DC precursor potential of a spleen was first determined. A suspension of spleen cells, with dead cells and erythrocytes removed, was counted and injected intravenously, and the number of donor-type CD8α⁺ DCs arising after 3.5 days was determined. We calculated the per-cell–specific precursor activity and then multiplied this by the number of viable nucleated cells obtained from each donor spleen to give an initial total precursor activity figure for balance sheet purposes (Table 2).

The first step in DC purification involved separating a light-density (less than 1.077 g/cm³) fraction, which included only approximately 5% of the viable nucleated spleen cells but virtually all of the fully developed DCs. Although this fraction gave some enrichment of CD8α⁺ DC precursor–specific activity (approximately 5-fold), it was less than the 20-fold obtained for DCs. More important, the balance sheet of CD8α⁺ DC total precursor activity showed that 70% was eliminated with the dense cells (Table 2). Additional studies showed that most of the immediate precursors of CD8α⁺ DCs were within the 28% of spleen viable nucleated cells, with density ranging between 1.083g/cm³ and 1.077 g/cm³.

Further enrichment of fully developed DCs, by successive immunomagnetic bead depletion of non-DC lineage cells and sorting to remove autofluorescent macrophages, enriched/DC purity to 70% to 80% but gave little further enrichment of CD8⁺ DC precursor–specific activity (Table 2). The final DC-enriched fraction was only 5-fold enriched in specific CD8α⁺ precursor activity, and it represented only 11% of the total precursor activity in the original spleen suspension. Much of this apparent activity could have resulted from the seeding and persistence of preformed CD8α⁺ DCs rather than their generation from precursors. Thus, most of the splenic CD8α⁺ DC precursors were not CD8α^- DCs.

Because some apparent CD8α⁺ DC precursor activity did persist in the DC-enriched fraction, we asked whether CD8α^- DCs could be a real, if minor, source of CD8α⁺ DCs. The enriched fraction (70%-80% DCs) was therefore segregated according to surface expression of the 3 markers used to define CD8α^- DCs—CD11c, MHC 2, and CD8α. Labeling the DCs with antibodies to these markers did not result in any loss of CD8α⁺ DC precursor activity (Table 2). On sorting the labeled cells, precursor activity was found distributed over many fractions (Table 3), not just in the CD8α^-, CD11c^hi, or MHC 2^hi fractions. On separation by CD8α expression levels, some “activity” was associated with the CD8α high fraction, and this presumably represented the seeding and persistence of pre-existing CD8α⁺ DCs. Even more of the total and specific activity was found in the CD8α intermediate fraction, however, suggesting the presence of less mature cells en route to producing CD8α high DCs (Table 3). On separation according to CD11c or MHC 2 expression, the greatest enrichment of activity was in the fractions expressing levels of CD11c and MHC 2 that were lower than those of fully developed DCs and that might therefore contain early DC forms (Table 3). In no case was the precursor activity enriched along with the markers of CD8α^- DCs (MHC 2^hi, CD11c^hi, or CD8α^-), though some measurable activity was usually associated with these markers, as in Figure 1. When all 3 markers were used together to isolate CD8α^-CD11c^hiMHC 2^hi DCs (as in Figure 1), a mean of 1.1% of the total initial spleen precursor activity was recovered in the 3 of 7 experiments when any CD8α⁺ DCs were found in the recipient spleens (data not shown).

In at least some of our experiments, CD8α⁺CD205⁺ splenic DCs were generated when purified samples of CD8α^- DCs were transferred intravenously, in accordance with the observations of Martinez del Hoyo et al.¹²  However, our conclusions differ radically from theirs because we demonstrate that most immediate precursors of CD8α⁺ DCs are not CD8α^- DCs at all but are cells lacking the characteristics of mature DCs and so are lost on DC enrichment. Our results highlight the importance of a balance sheet approach, examining activity at all enrichment steps rather than only determining the activity of an enriched candidate cell.

A major limitation of the earlier study was the extremely low recovery of DCs in the spleen following intravenous transfer. We also found this with at most only 0.16% of the sorted CD8α^- DCs transferred, forming CD8α⁺ DCs in the recipient spleens. This poor recovery of DCs after intravenous transfer is compounded by the separate problem of poor recovery of CD8α^- DC precursor activity during DC enrichment and purification. When we did recover any CD8α⁺ DC precursor activity in a highly purified, sorted CD8α^-CD11c^hiMHC 2^hi fraction, it represented only 1.1% of the total initial spleen activity. Even allowing for some DC loss during enrichment, this is a very low recovery of activity. However, this final result is in line with the results in Tables 2 and 3, indicating that most CD8α⁺ DC precursor activity did not segregate with individual markers for the CD8α^- fully developed DCs. Although this low-level production from the sorted CD8α^- DCs might point to one minor route of CD8α⁺ DC production, it could also arise from a high-efficiency precursor contaminant. Rather than resolve this issue, we consider future investigation should focus on the major pathways leading to CD8α⁺ (and CD8α^-) DCs.

What, then, are the characteristics of most immediate precursors of the CD8α⁺ DCs found in normal laboratory mice? We find they are lighter in density than most spleen lymphocytes but more dense than mature DCs, but at present we have little further information. One candidate precursor, the mouse plasmacytoid cell, has recently been eliminated because it produces a subtype of CD8α⁺ DCs that differs in other markers and that only produces this CD8α⁺ DC subtype after microbial stimulation.¹⁵  Our kinetic studies^9,15  indicate that the 3 mature DC subsets and the plasmacytoid cells of normal mouse spleen represent separate development streams at least as far back as the last dividing precursor. However, this may not be very far back, and the separate streams could well branch off from a common DC-committed precursor.¹⁶  The relationship between all these DC subtypes and the nature and lineage commitment of their immediate precursors remain to be elucidated.