More on the differentiation of dendritic cells into osteoclasts

Response

Alnaeeli and Teng raised questions regarding our recent publication on the differentiation of dendritic cells (DCs) into osteoclasts (OCLs).¹  Despite the fact that they have themselves purified DCs using CD11c as an exclusive marker,²  they raised concerns about its specificity. Absence of contaminant cells was our major concern and led us to develop a protocol for reaching the highest purification rate.¹  Furthermore, cDCs were purified using not only CD11c but coexpression of CD11c and MHC-II, specific of DCs.³  Phenotype of the purified CD11c^highMHC-II⁺ cells and the use of CD11c-DTR-mice confirmed the absence of contamination (Wakkach et al, Figures 1 and 3A,B¹ ). Finally, our data are strongly supported by other reports, including from Alnaeeli's group.^2,4,5  Therefore, we strongly believe that the effects observed in vivo are due to cDCs and not to other cell types.

We did not state that purified-cDCs are “100%-pure-mature-DCs.” Indeed, despite splenic cDCs expressing some levels of maturation markers, they can be further matured with lipopolysaccharide (LPS),⁶  and therefore they are neither fully mature nor comparable to in vitro–generated immature DCs. OCL differentiation from LPS-activated DCs is evidenced in Figure S1C.¹  Inconsistency with previous studies^2,5  and results presented by Alnaeeli and Teng is probably due to completely different conditions used for DC generation and maturation. Moreover, our result is not in contradiction with apoptosis of mature DCs because we have treated BM-derived LPS-activated DCs with RANK-L, a potent inhibitor of DC apoptosis shown to efficiently protect DCs from apoptosis at least for 6 days in culture.⁷ 

We state that 80% of the cells present in vitro at the end of the differentiation are multinucleated TRAP⁺. Formal quantification was not done because similar efficiency was previously reported, including by Alnaeeli et al (95% multinucleated TRAP⁺ cells).^2,4  The fact that such a high efficiency could be due to contamination could therefore be a concern for all these studies. However, we do not believe so because “classical” OCL-precursors are less efficient than DCs regarding osteoclastogenesis.⁴ 

In vivo, we did not quantify cDC-derived OCLs because they were rare due to the few injected DCs reaching the bone marrow (Wakkach et al, Figure 4A¹ ), explaining the low resorptive activity observed in vitro and in vivo (Wakkach et al, Figures 3 and 4D¹ ). Instead, we used recipient mice devoid of OCL activity, the osteopetrotic oc/oc mice, that allow easy detection of few active OCLs through partial rescue of their bone phenotype. Considering the efficiency of phenotype rescue reported for hematopoietic cell transplantation in oc/oc mice,^8-10  our results are indeed spectacular and it would have been highly surprising to have a higher recovery rate.

DC-derived OCLs have been detected at the bone surface (Wakkach et al, Figure 4C¹ ) using the a3 “internal” marker instead of green fluorescent protein/carboxyfluorescein succinimidyl ester, because a3 is not expressed in recipient oc/oc mice. Therefore, a3 expression is sufficient to specifically identify injected DCs and OCLs deriving from them. The weak a3 expression by these OCLs probably results from the fusion of endogenous-a3^neg–pre-OCLs with a3⁺ injected DCs, leading to a dilution of the signal.

Regarding RANK-L expression, although it was analyzed by reverse-transcribed–polymerase chain reaction, we believe that this does not decrease the importance of our findings. Our hypothesis is not in contradiction with that of Alnaeeli et-al²  because it also supposes a contact between CD4⁺ T cells and DCs for activation and polarization. However, we agree that understanding the contribution of inflammatory CD4⁺ T cells to the differentiation of DCs into OCLs needs further analysis, in particular in the context of inflammatory osteolysis.