Maintenance of memory T cells in the bone marrow: survival or homeostatic proliferation?

Six years ago, Radbruch and colleagues discussed in Nature Reviews Immunology (Organization of immunological memory by bone marrow stroma. Nat. Rev. Immunol. 10, 193–200 (2010))¹ how distinct stromal cell subsets in the bone marrow can support the life-long persistence of plasma cells and memory T cells. These authors proposed that the bone marrow might serve as a depot for resting non-circulating memory T cells. Furthermore, they discussed how memory T cells might be maintained in the bone marrow by survival factors, such as interleukin-7 (IL-7), as opposed to by proliferative factors, such as IL-15. This view was in contrast with the largely accepted notion at the time that recirculating memory T cells are maintained by a homeostatic equilibrium between proliferation and death long after antigen clearance². Furthermore, it did not accommodate previous data concerning the proliferation^3,4 and recirculation^5,6 of memory CD8⁺ T cells in the bone marrow.

Recently, the idea that was originally proposed in Nature Reviews Immunology¹ was revived by the identification of quiescent, non-migratory tissue-resident memory T (T_RM) cells in the skin, gut and other organs⁷. Indeed, Radbruch et al. hypothesized that bone marrow memory T cells might share several features with T_RM cells, and they suggested that their previous and newly generated findings supported this concept^8,9. However, it might be misleading to chiefly consider bone marrow memory T cells as non-circulating, non-dividing cells.

Experiments using Ki67 staining in mice and humans have shown that, at any given time-point, 95–98% of memory CD8⁺ T cells in the bone marrow are in the G0 phase of the cell cycle^8,9. Of the remaining cells, some are in the G1 interval, and a few (that is, 0.2–1.7%) are actively proliferating in S/G2/M^3,8,9. However, this still means that the proportion of memory CD8⁺ T cells proliferating in the bone marrow is reproducibly two- to fourfold higher than the proportions (that is, 0.05–0.80%) proliferating in the spleen, lymph nodes or blood^3,8,9. This is true also when cell division is measured for one or more days. For example, in a 3 day-bromodeoxyuridine (BrdU)-labelling analysis, the average frequency of dividing antigen-specific memory CD8⁺ T cells was 4% in the bone marrow and 2% in the spleen⁴. Moreover, when carboxyfluorescein succinimidyl ester (CFSE)-labelled antigen-specific memory CD8⁺ T cells were transferred into non-immunized animals, they showed higher rates of proliferation in the bone marrow than in the spleen and lymph nodes^3,10. In general, the data concerning memory CD8⁺ T cell cycling in the bone marrow are all in agreement. However, memory CD4⁺ T cell proliferation requires further investigation, as antigen-specific cells have not been examined in the bone marrow¹¹.

Despite the apparent consistency of the data concerning memory CD8⁺ T cells, their interpretations differ. Radbruch's group proposes that these data suggest similarity between bone marrow memory CD8⁺ T cells and peripheral T_RM cells, reinforcing the concept that a resting non-proliferative state following antigen clearance is the hallmark of memory CD8⁺ T cells^8,9. These authors suggest that the number of memory CD8⁺ T cells proliferating in the bone marrow is negligible and may have been overestimated owing to a BrdU-related artefact⁹. They also showed by RNA microarray analysis that bone marrow memory CD8⁺ T cells resembled their spleen counterparts and that both had overtly different transcriptomes from memory CD8⁺ T cells stimulated in vitro⁹. Moreover, they have suggested that bone marrow memory CD8⁺ T cells are sessile, as up to 60% of them express CD69, a molecule that in CD4⁺ T cells is essential for retention in the bone marrow^8,9. Finally, they suggest that the colocalization of bone marrow memory CD8⁺ T cells with IL-7-producing stromal cells supports the idea of IL-7-driven survival of memory CD8⁺ T cells in the absence of proliferation⁹.

However, it could be argued that although the frequency of memory CD8⁺ T cells that are dividing in the bone marrow is low, the absolute numbers of proliferating memory CD8⁺ T cells is much higher in the bone marrow than in the spleen and lymph nodes^3,4. As regards BrdU-related artefacts, they may occur at high BrdU doses⁹ but seem uncommon at the standard BrdU dose that was used in bone marrow T cell studies^3,4,12. Notably, recent adoptive transfer experiments in genetically modified mice have shown that IL-15 in the bone marrow promotes proliferation and inhibits interleukin-7 receptor subunit-α (IL-7Rα) expression in memory CD8⁺ T cells, independently of antigen co-transfer or treatment with innate receptor agonists¹³. In respect to molecular data⁹, it is perhaps not surprising that transcription profiles were highly diverse when comparing ex vivo-isolated and in vitro-stimulated cells; besides, some differences between freshly obtained spleen and bone marrow CD44^hiCD8⁺CD3⁺ T cells might have been missed owing to the cell sorting strategy. For instance, IL-7Rα^hi, but not IL-7Rα^low, T cells were selected for analysis, and yet IL-7Rα^low T cells are enriched in the bone marrow^8,13, reflecting in vivo exposure to IL-15 (Ref. 13). Moreover, global transcription data, CD69 expression profiles and colocalization in tissue sections do not address in vivo T cell migration. In fact, in situ-labelling studies and parabiosis experiments have shown that memory T cells do recirculate to and from the bone marrow^5,6.

In conclusion, the available evidence supports the view that the bone marrow is a 'stopping point' where recirculating memory CD8⁺ T cells are stimulated to proliferate before continuing to move around the body^3,4,12,13. Notably, lodging into the bone marrow is a competitive process among memory T cells¹⁴, which is an element to be considered especially in interpretation of adoptive transfer data^11,14. Furthermore, reported diversities in the repertoire of antigen specificity between human bone marrow and blood memory CD4⁺ T cells after in vitro restimulation⁸ might reflect several features, including: differences in T cell recruitment into the bone marrow, undetected ongoing responses against common pathogens (for example, Candida albicans and Cytomegalovirus) and differences between in vitro and in vivo responses. Finally, recent parabiosis experiments demonstrated that a small percentage (up to 5%) of memory CD8⁺ T cells in the spleen and lymph nodes are non-migratory¹⁵. Therefore, it is possible that a minor population of memory CD8⁺ T cells in the bone marrow (as opposed to the majority) might be sessile, but direct evidence for this is lacking at the moment.