The origins of memory T cells

Memory T cells protect against previously encountered pathogens, but their origins are unclear. Two studies track DNA modifications over time and find that these cells arise from effector T cells.
Kyla D. Omilusik is in the Division of Biological Sciences, University of California, San Diego, La Jolla, California 92093, USA.

Search for this author in:

Ananda W. Goldrath is in the Division of Biological Sciences, University of California, San Diego, La Jolla, California 92093, USA.

Search for this author in:

Vaccination has substantially reduced illness and death from infectious disease by exploiting the ability of long-lived memory T cells to ‘remember’ a previous encounter with a specific microbe and mount a rapid response upon pathogen re-exposure. Understanding how immunological memory is established and maintained might provide insights that could enable improvements in vaccine design. In this issue, Akondy et al.1 (page 362) and Youngblood et al.2 (page 404) reveal the cell population that gives rise to memory T cells, and how the population of memory T cells evolves.

Naive T cells are those that haven’t previously responded to a pathogen. When they recognize a pathogen, they rapidly divide and express molecules such as cytokine proteins that help to fight infection. These responding cells are called effector T cells (more specifically, a type of effector cell called a cytotoxic T cell) and they can migrate into inflamed tissues and kill infected cells3. Once the pathogen is eliminated, most effector cells die, but a small pool of long-lived memory cells remains that is poised to respond rapidly if reinfection occurs3.

Which cells give rise to memory T cells has been extensively investigated. Two general possibilities (Fig. 1a, b) have been proposed: the cells either arise from a subset of the effector cells that escape death, or instead descend directly from naive T cells, which could, as early as their first cell division, give rise to cells with effector-T-cell or memory-T-cell potential3.

Figure 1 | The formation of memory T cells. When naive T cells encounter a pathogen, they proliferate, giving rise to many effector T cells that fight infection and die when the infection is over. Pathogen infection also results in the formation of long-lived memory T cells that rapidly respond to a previously encountered pathogen. a, One model for memory-T-cell formation proposes that the precursor cells that give rise to memory T cells and effector T cells both arise independently from naive T cells. b, In another model, a subset of effector cells gives rise to memory T cells. c, Akondy et al.1 and Youngblood et al.2 studied memory-T-cell formation in human T cells and in vivo in mice, respectively. The authors monitored the addition of methyl (Me) groups to DNA, a change usually associated with a decrease in gene expression. They analysed the methylation of genes associated with naive-cell function and effector-cell function in naive cells, effector cells and memory cells to reveal the cells’ developmental histories. The authors also analysed the patterns of DNA accessibility (not shown) in the different cell types. They find that effector-function genes have similar methylation patterns in effector T cells and memory T cells, which could allow memory T cells to rapidly re-express (dotted arrow) those genes upon pathogen reinfection. Their evidence is consistent with the model shown in b.

The two new studies aimed to resolve this debate by tracking CD8+ T cells (those that express the protein CD8 on their cell surface) during the course of an infection. Akondy et al.1 studied people who received a vaccine for yellow fever virus, whereas Youngblood et al.2 worked with a mouse model of lymphocytic choriomeningitis viral infection. Both studies examined populations of naive, effector and memory T cells for epigenetic modifications — heritable chemical modifications or structural changes to a cell’s DNA that do not alter the DNA sequence. Such changes are often associated with the regulation of gene expression, allowing a cell and its descendants to ‘bookmark’ a particular gene-expression pattern. DNA methylation is one such alteration that can fix genes in an ‘off’ position, thus silencing their expression4.

Akondy and colleagues also examined the DNA configuration in T cells to identify genomic regions tightly packaged in a ‘closed’ state, which are not accessible to the gene-expression machinery, and regions in an ‘open’ state that allows gene expression5. Although RNA profiling can provide a snapshot of the genes currently being transcribed by a cell, the approaches taken by the authors to analyse epigenetic changes provide insight into the path by which a given transcriptional state is reached.

Using genome-wide DNA-methylation profiling, Youngblood and colleagues found that, as naive T cells differentiated into effector cells, their DNA-methylation profile changed. Methyl groups were added to many genes associated with the naive state, whereas a loss of DNA methylation was observed at genes that encode key components of the effector response.

The authors identified the DNA methyltransferase Dnmt3a as a key enzyme responsible for de novo DNA methylation during the immune response. They also showed that, in effector T cells that are differentiating into memory cells, the methylation of genes expressed in the naive state can be removed in a cell-division-independent process that might drive the re-expression of naive-associated genes, which could, in turn, be needed to establish or maintain the long-lived memory-cell population.

Importantly, both studies conclude that, although memory T cells no longer express effector molecules, the genes that encode these molecules remain in a state of low methylation (Fig. 1c). Akondy and colleagues also reveal that memory T cells that are present as long as ten years after vaccination have an open genome-packaging pattern that is comparable to that found in effector T cells, even though these memory T cells do not divide or express effector molecules until reactivated by pathogen encounter. Thus, effector-associated genes were maintained in memory T cells in a configuration that more closely resembled that of effector T cells than that of naive-T-cell populations. This is consistent with the ability of memory cells to re-express effector molecules quickly to fight reinfection3.

Both studies provide strong evidence for a model in which memory T cells descend from a population of cells that have previously expressed genes associated with effector-T-cell function, and have turned off effector-gene expression but retained a ‘memory’ of their developmental path. This takes the form of DNA modifications that would enable them to rapidly become an effector cell upon pathogen reinfection. In essence, epigenetic modifications bookmark the chapter of infection in a memory T cell’s history, allowing the cell to rapidly ‘recall’ its effector capacity.

Although inferring T-cell lineages by isolating bulk populations and comparing heritable epigenetic states is illuminating, it does not exclude the possibility that a small population of naive T cells could also independently form memory cells directly — without going through an effector-T-cell phase. Akondy and colleagues addressed this possibility by directly labelling dividing effector T cells in individuals using a heavy form of hydrogen called deuterium.

Virus-specific memory T cells that were present between one and two years after vaccination showed little dilution of the deuterium, indicating that minimal cell division had occurred in these cells. In a comparison of cell-surface proteins, these memory cells resembled naive T cells, but the deuterium labelling reveals that they were formed from the dividing effector-T-cell population. These results support a model in which viral-specific CD8+ T cells extensively proliferate upon pathogen recognition and modify their DNA to favour expression of effector molecules. Later the cells stop dividing, stop expressing effector genes and re-express many genes associated with the naive state, such as genes that aid T-cell survival and migration.

Not all memory T cells are the same: a proportion of them, like naive cells, circulate throughout the body waiting to be called back into action, whereas tissue-resident memory cells reside in tissues such as the lung, skin and gut, providing the first line of defence against reinfection at sites where pathogens will often first enter the body6. Whether these different types of memory T cell arise through the same pathway remains to be determined.

These two studies suggest that a goal of vaccine design should be to stimulate a large, robust response from the effector T cells from which memory T-cell populations can arise. However, the conditions that best promote effector cells to become memory cells still need to be fully defined. Therapeutic targeting of the DNA-modification machinery, such as Dnmt3a, might prove to be a useful strategy to increase vaccine efficacy.

Nature 552, 337-339 (2017)

doi: 10.1038/d41586-017-08280-8


  1. 1.

    Akondy, R. S. et al. Nature 552, 362–367 (2017).

  2. 2.

    Youngblood, B. et al. Nature 552, 404–409 (2017).

  3. 3.

    Kaech, S. M. & Cui, W. Nature Rev. Immunol. 12, 749–761 (2012).

  4. 4.

    Winter, D. R., Jung, S. & Amit, I. Nature Rev. Immunol. 15, 585–594 (2015).

  5. 5.

    Gray, S. M., Kaech, S. M. & Staron, M. M. Immunol. Rev. 261, 157–168 (2014).

  6. 6.

    Chang, J. T., Wherry, E. J. & Goldrath, A. W. Nature Immunol. 15, 1104–1115 (2014).

Download references

Nature Briefing

An essential round-up of science news, opinion and analysis, delivered to your inbox every weekday.