Human memory T cells from the bone marrow areresting and maintain long-lasting systemic memoryAnna Okhrimenkoa, Joachim R. Grünb, Kerstin Westendorfa, Zhuo Fanga,c, Simon Reinked, Philipp von Rothe,Georgi Wassilewe, Anja A. Kühlf, Robert Kudernatschg, Sonya Demskig, Carmen Scheibenbogeng, Koji Tokoyodah,Mairi A. McGratha, Martin J. Rafteryi, Günther Schönrichi, Alessandro Serraa, Hyun-Dong Changa, Andreas Radbrucha,1,and Jun Donga,1,2

aCell Biology, bBioinformatics, cSignal Transduction, and hOsteoimmunology, Deutsches Rheuma-Forschungszentrum Berlin, 10117 Berlin, Germany;dCore Unit Cell Harvesting, Berlin-Brandenburg Center for Regenerative Therapies and gClinical Tumor Immunology and Immunemonitoring, Institutefor Medical Immunology, Charité University of Medicine Berlin, 13353 Berlin, Germany; eCenter for Musculoskeletal Surgery and iInstitute of Virology,Charité University of Medicine Berlin, 10117 Berlin, Germany; and fDepartment of Gastroenterology, Infectiology and Rheumatology, Charité Universityof Medicine Berlin, 12200 Berlin, Germany

Edited by Klaus Rajewsky, Max Delbrück Center for Molecular Medicine, Berlin, Germany, and approved May 16, 2014 (received for review October 3, 2013)

In the bone marrow, a population of memory T cells has beendescribed that promotes efficient secondary immune responsesand has been considered to be preactivated, owing to its expressionof CD69 and CD25. Here we show that human bone marrowprofessional memory T cells are not activated but are resting interms of proliferation, transcription, and mobility. They are in theG0 phase of the cell cycle, and their transcriptome is that of restingT cells. The repertoire of CD4+ bone marrow memory T cells com-pared with CD4+ memory T cells from the blood is significantlyenriched for T cells specific for cytomegalovirus-pp65 (immunodo-minant protein), tetanus toxoid, measles, mumps, and rubella. Itis not enriched for vaccinia virus and Candida albicans-MP65(immunodominant protein), typical pathogens of skin and/ormucosa. CD4+ memory T cells specific for measles are maintainednearly exclusively in the bone marrow. Thus, CD4+ memory T cellsfrom the bone marrow provide long-term memory for systemicpathogens.

antigen-specific response | short- and long-term memory | polyfunctional

Memory CD4+ and CD8+ T cells are essential componentsof immunological memory. In humans, analyses of memory

T cells have been largely limited to T cells isolated from peripheralblood. Memory T cells of the blood and secondary lymphoidorgans (SLOs) are believed to recirculate to find their cognateantigen, presented to them by activated antigen-presenting cells(APCs) in the SLOs. In the apparent absence of antigen theirnumbers slowly decline (1, 2), suggesting that recirculating mem-ory T cells provide a memory of a limited duration. They havebeen described as a heterogeneous population with respect tofunction, based on the expression of either chemokine receptors(3) or cytokines (4). With respect to immunity against infectiouspathogens such as HIV, polyfunctional memory Th cells areconsidered to be the most effective in protection (5).Recently, memory T cells have been found in many peripheral

tissues, such as skin, lung, intestine, and thymus (6–8). Thesetissue-resident memory T cells (TRM) are thought to provideimmediate effector function at the preferred sites of infectionwith specific pathogens; however, the stability of this memory inhumans is not yet clear (9). Memory T cells have also beenidentified in the bone marrow, which is a privileged tissue in thatit is not connected to lymphatic vessels. In humans, memoryT cells from the bone marrow express CD69 and thus have beenconsidered to be “preactivated” (10) and, until now, their con-tribution to immunological memory has remained unclear.In mice, we have recently identified a population of professional

memory CD4+ T cells that is located in the bone marrow in thememory phase of an immune response (1), similar to long-livedplasma cells (11–13). These memory Th cells dock onto IL-7–expressing stroma cells (1), which presumably provide a survival

niche for their maintenance (14–16), analogous to the maintenanceof long-lived plasma cells by CXCL12-expressing stroma cells (11–13). Murine memory Th cells of the bone marrow provide efficienthelp to B cells in secondary immune reactions, and their numbersare stably maintained. Furthermore, murine bone marrow memoryTh cells express CD69, which is critically required for migration ofTh cells into the bone marrow (17). Most importantly, in the ab-sence of bone marrow memory Th cells, secondary immuneresponses are severely impaired, demonstrating the essential roleof murine bone marrow memory Th cells for the maintenance ofimmunological memory.Here we analyze and compare memory T lymphocytes from

human bone marrow and blood. We show that memory T cells ofthe bone marrow rest in terms of proliferation, transcription, andmobility, despite expressing CD69. The repertoire of bone marrowCD4+ memory T cells is enriched for T cells recognizing cyto-megalovirus (CMV)-pp65, tetanus toxoid (TT), rubella, mumps,and measles virus, all of which are systemic pathogens. CD4+

memory T cells specific for vaccinia virus or Candida (C.) albicans-MP65, representing skin- and/or mucosa-specific pathogens, are

Significance

Memory T cells are essential components of immunologicalmemory. In the apparent absence of antigen, numbers of recir-culating antigen-specific memory T cells dwindle, provoking thequestion of whether there is immunological memory withoutmemory T cells. Here we show that human memory T cells canreside in the bone marrow as resting cells in terms of pro-liferation, transcription, and mobility. The repertoire of bonemarrow memory T cells is enriched for systemic pathogensrepresenting persistent, recent, and childhood challenges. Interms of absolute numbers, memory T cells specific for systemicantigens are maintained predominantly in the bone marrow, inparticular those representing historic encounters.

Author contributions: A.R. and J.D. designed research; A.O., K.W., and J.D. performedresearch; S.R., P.v.R., G.W., A.A.K., R.K., S.D., C.S., K.T., M.J.R., and G.S. contributed newreagents/analytic tools; A.S. and H.-D.C. wrote the ethical application; A.O., J.R.G., Z.F.,M.A.M., and J.D. analyzed data; A.O. and J.D. interpreted data and generated figures;and A.R. and J.D. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Data deposition: The data discussed in this paper have been deposited in the GeneExpression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no.GSE50677).1A.R. and J.D. contributed equally to this work.2To whom correspondence should be addressed. E-mail: dong@drfz.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1318731111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1318731111 PNAS | June 24, 2014 | vol. 111 | no. 25 | 9229–9234

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 28

, 202

0

not enriched in the bone marrow. In terms of absolute numbers,most memory CD4+ T cells specific for systemic antigens aremaintained in the bone marrow, in particular those specific forhistoric antigens. Memory CD4+ T cells of the bone marrow arepolyfunctional, namely they have the capability to express severalcytokines simultaneously. Thus, memory T cells of human bonemarrow are resting, professional memory cells that provide long-lasting, polyfunctional memory for systemic pathogens.

ResultsHuman Memory T Cells in Bone Marrow and Peripheral Blood. Wedetermined the frequencies of CD3+CD4+CD45RO+CD45RA−

and CD3+CD8+CD45RO+CD45RA− bona fide memory T cellsamong all CD3+ T lymphocytes from 12 paired bone marrow andblood samples of 50- to 60-y-old individuals by flow cytometry. Inblood, 52% of CD3+CD4+ and 36% of CD3+CD8+ T cells wereCD45RO+CD45RA− memory cells; in bone marrow, 51% ofCD3+CD4+ and 42% of CD3+CD8+ T cells were CD45RO+

CD45RA− memory cells (Fig. 1A). Of the CD3+ T cells in blood,73% were CD4+ and 19% were CD8+, whereas in bone marrow,54% were CD4+ and 32% were CD8+ (Fig. 1B). For both CD4+

and CD8+ T cells, their distributions in blood and bone marrowwere likewise consistent between analyzed donors (Fig. 1 A andB). Based on the reasonable estimate that the number of T cellsin bone marrow is 25 × 109 and that of blood is 7 × 109 (18, 19),human bone marrow thus contains three to four times morememory T lymphocytes than blood (Fig. 1C), roughly 10 × 109

cells. In addition to CD45RO+CD45RA− memory T cells, bloodand bone marrow also contain CD45RA+CD45RO−CCR7− mem-ory T cells (TEMRA) (5 × 108 and 3 × 109, respectively) (Fig. S1 Aand B). Among CD4+ T cells, TEMRA from bone marrow and bloodwere 25 and 29 times less frequent than CD45RO+CD45RA−

memory cells, respectively, whereas among CD8+ T cells, TEMRAwere as similarly frequent as CD45RO+CD45RA− cells in bloodand bone marrow.

Human Bone Marrow Memory T Cells Express CD69 but Are NotActivated. CD69 is a calcium-dependent type II transmembranereceptor of the lectin superfamily (20). Expression of CD69is induced upon activation of T lymphocytes and is thereforesometimes regarded as an activation marker (21). Less than 1%of human CD3+CD45RO+CD45RA− memory T cells from pe-ripheral blood expressed CD69. Strikingly, in bone marrow,62.8% of CD8+ and 28.6% of CD4+ memory T cells expressedCD69, whereas in terms of absolute numbers, bone marrowcontained equal numbers of CD69+ memory CD4+ and CD8+ Tcells (Fig. 1D). With the exception of CD69, bone marrowmemory T cells did not express other putative activation mark-ers. For example, ex vivo isolated memory T cells from blood orbone marrow did not express CD137 (Fig. 1E). This was not dueto anergy or refraction, because memory T cells from blood andbone marrow readily expressed CD137 upon restimulation withanti-CD3/CD28 (Fig. S2). CD25 was expressed by 9% of the(CD69−) CD4+ cells from blood and by 2.1 and 5.5% of the bonemarrow CD69+ and CD69− memory CD4+ T cells, respectively(Fig. 1F). However, more than 80% of the cells expressing CD25did not express the IL-7 receptor (CD127low) but rather expressedFOXP3 (forkhead box P3; scurfin) (Fig. 1G and Fig. S3), high-lighting that these cells are bona fide regulatory T cells (22).Therefore, bone marrow also contains a substantial fraction ofregulatory memory T cells and, moreover, bona fide memory Tcells residing in the bone marrow are not in an activated state,despite their expression of CD69.

Human Bone Marrow Memory T Cells Rest in Terms of Proliferation,Transcription, and Mobility. To analyze steady-state proliferationof memory T cells in paired bone marrow and blood samples atthe single-cell level, cells were stained for expression of Ki67, an

antigen that is expressed in all phases of the cell cycle except G0(23). On average, 4.2% of memory CD4+ and 5.0% of memoryCD8+ T cells isolated ex vivo from blood expressed Ki67. In thebone marrow of the same donors, 1.2% of CD4+ and 1.7% ofCD8+ memory T cells expressed Ki67. Some of those Ki67+ cellsof blood and bone marrow in all likelihood represent recentlygenerated effector cells, because they express only low levels of

% o

f Max

% o

f Max

CD69

D

Mem

ory

CD

4+ T

cel

lsM

emor

y C

D8+

T c

ells

C

BMPBC

D69

CD137

Memory CD4+ T cells Memory CD8+ T cells

F G%

of M

ax

CD127 FOXP3

E

PBBM

PBBM

PB CD25hiBM CD25+CD69+BM CD25+CD69-PB CD25-

1.0 0.1

1.1

0.9

0.7

30.7 2.3 0.1

0.7

1.8

0.3

62.3

0

5

10

15 p = 0.0005

Mem

ory

CD

4+T

cells

( x10

e9)

PB BM0

2

4

6 p = 0.0005

Mem

ory

CD

8+T

cells

(x10

e9)

0

1

2

3

4 p = 0.0005

CD

69+ce

lls(x

10e9

)

PB BM0

1

2

3

4 p = 0.0005

CD

69+ce

lls(x

10e9

)

0

10

20

30

40 p < 0.0005

CD

69+

cells

(%)

PB BM0

20

40

60

80 p < 0.0005

CD

69+

cells

(%)

PB BM0369

1215 CD69-

CD69+

CD

25+

cells

(%)

PB BM

BA

PB BM0

20406080

100p = 0.0005

CD

4+T

cells

(%)

PB BM0

20

40

60

80p = 0.0025

CD8+

T ce

lls (%

)

PB BM0

20406080

100 p = 0.8445

Mem

oryC

D4+ T

cells

(%)

PB BM0

20

40

60

80 p = 0.0210

Mem

ory

CD

8+T

cells

(%)

Fig. 1. Memory T cells in human bone marrow and peripheral blood. (A andB) Frequencies of CD45RO+CD45RA− memory CD4+ and CD8+ T cells (A) andfrequencies of CD4+ and CD8+ T cells (B) from paired bone marrow (BM)and peripheral blood (PB) samples. (C) Absolute numbers were calculatedaccording to the frequencies of memory CD4+ and CD8+ T cells (A) and ofCD4+ and CD8+ T cells (B) and according to the estimated numbers of T cellswithin BM and PB of healthy young adults (18, 19). (D) Memory CD4+ andCD8+ T cells of BM express CD69. A representative plot of overlaid histo-grams of CD69 expression on memory CD4+ or CD8+ T cells of BM and PB,and the average frequencies and absolute numbers of CD69+ memory CD4+

and CD8+ T cells are shown. In A–D, n = 12; P values were obtained as de-scribed in SI Materials and Methods. (E) CD137 expression on CD69+ andCD69− memory CD4+ and CD8+ T cells from BM and PB samples. Data shownare representative of four separate experiments. (F) Average frequencies ofCD25+ cells among CD69+ or CD69− memory CD4+ T cells of five paired BMand PB samples. Error bars represent SEM. (G) CD127 and FOXP3 expressionon CD69+CD25+ and CD69−CD25+ memory CD4+ T cells of BM and onCD25hiCD69− and CD25−CD69− memory CD4+ T cells of PB. Data shown arerepresentative of five separate experiments.

9230 | www.pnas.org/cgi/doi/10.1073/pnas.1318731111 Okhrimenko et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 28

, 202

0

CD127 (24) (Fig. S4). Thus, more than 98% of the bone marrowmemory T cells are resting in the G0 phase of the cell cycle (Fig.2A). In relation to CD69 expression by memory CD4+ and CD8+

T cells of the bone marrow, Ki67 expression was largely confinedto CD69− memory CD4+ and CD8+ T cells (Fig. 2B). To furtheridentify memory T cells in the S and G2/M phases of the cellcycle, DNA content was assessed quantitatively by propidiumiodide (PI) staining (Fig. S5). On average, 0.185% of memoryCD4+ cells from blood and 0.175% from bone marrow, and0.067% of memory CD8+ T cells from blood and 0.2% frombone marrow, were in the S and G2/M phases of the cell cycle(Fig. 2C). Importantly, memory T cells from both blood andbone marrow could be readily reactivated with anti-CD3/CD28,resulting in 15–22% of the CD4+ and CD8+ memory T cellsentering the S and G2/M phases of the cell cycle (Fig. 2C and Fig.S5). Therefore, steady-state memory T cells of human bonemarrow are resting in the G0 phase of the cell cycle.To further determine the global transcriptional activity of bone

marrow memory CD4+ T cells, gene expression was analyzed byAffymetrix microarrays. Memory CD4+ T cells were isolated exvivo from four paired blood and bone marrow samples. Tran-scriptomes of these cells were compared with those of memoryCD4+ T cells from blood of eight unrelated donors, directly afterex vivo isolation or after stimulation for 3 h with phorbol-myristicacid (PMA)/ionomycin. Transcriptomes of memory CD4+ T cellsfrom blood and bone marrow were very similar and clearly distinctfrom transcriptomes of activated memory CD4+ T cells (Fig. 3). Inmemory CD4+ T cells from blood and bone marrow, comparedwith activated T cells, genes indicating activation were down-regulated, as exemplified by IL22, FOXP1, and IFNGR2, whereasgenes such as RCOR3 and HDAC1, indicating repression oftranscription, were up-regulated (Table S1). Transcriptional in-activity of ex vivo cells was indicated by down-regulation ofIRF4, RNPS1, SLAMF1, and NFAT5, and proliferative rest wasindicated by down-regulation of MKI67IP, CDK6, and WEE1and up-regulation of RBL2. In addition, mobility arrest was in-dicated by up-regulation of ARHGDIB and down-regulation ofRHOA and SVIL. As expected, and of particular relevance, isthe selective down-regulation of S1PR1, a receptor for S1P, inCD69+ memory CD4+ T cells from bone marrow. This receptorwould be required for S1P-mediated egress into the blood (25–27). With respect to genes involved in survival, memory T cellsdid not express FASL and other apoptotic regulators such asCRADD and TNFAIP8, and thus are not prone to FAS-medi-ated induction of apoptosis, as expected, whereas activated Tcells did express FASL, CRADD, and TNFAIP8. In addition,expression of antiapoptotic regulators XAF1 and CD27 was up-regulated in the resting memory T cells, compared with the ac-tivated cells. With respect to functional diversity, expression of the

T-cell differentiation lineage-determining transcription factorsGATA3, TBX21 (TBET), RORC (nuclear receptor ROR-gamma), EOMES, and BCL6 was detected in memory T cellsfrom both blood and bone marrow, indicating a broad func-tional repertoire. Otherwise, the global gene expression profiles ofmemory T cells from blood and bone marrow confirm their restingstate in terms of proliferation and mobility, and are clearly distinctfrom activated cells.

The Repertoire of Bone Marrow Memory CD4+ T Cells. To assess therepertoire of bone marrow memory CD4+ T cells, we determinedthe frequencies of memory CD4+ T cells specific for CMV-pp65,TT, measles, rubella, mumps, vaccinia virus, and C. albicans-MP65 in paired bone marrow and blood samples of individualdonors aged 40–70 y. Bone marrow and blood mononuclear cellswere restimulated with antigen and CD28 antibodies for 12 or16 h (the latter for vaccinia virus). CD4+ T cells reacting to theantigen were identified according to the expression of CD154(28, 29) and one or more of the cytokines IL-2, TNF-α, and IFN-γ,as assessed by intracellular immunofluorescence (Fig. S6A). Fre-quencies of cells reacting to stimulation with “anti-CD28 only”were subtracted from the frequencies of cells reacting to “anti-CD28 plus antigen.” The superantigen SEB was included as a highcontrol for stimulations (Fig. S6B). Furthermore, to exclude thatany potential differences in frequencies of reactive cells were dueto inadequate antigen presentation in either blood or bone mar-row, for example different types or numbers of APCs, we per-formed control experiments mixing bone marrow and bloodcells at 1:1 ratios, with one or the other fraction being carboxy-fluorescein succinimidyl ester (CFSE)-labeled, and stimulatingthem with CMV-pp65. The frequencies of reactive cells from bonemarrow and blood were the same in the mixed culture as in theseparate pure blood and bone marrow control cultures, demon-strating that APCs from both bone marrow and blood are equallyefficient at presenting antigen (Fig. S6C).For pp65 and TT, CD154+cytokine+ cells were readily de-

tectable both in blood and bone marrow, but frequencies (Fig.4A) were higher in bone marrow than in blood, except for onedonor showing a higher frequency of TT-specific cells in bloodthan in bone marrow. Moreover, absolute numbers (Fig. 4B) ofpp65-reactive cells were significantly higher in bone marrow thanin blood. Similarly, TT-reactive cells in bone marrow out-numbered those in blood by five times (Fig. 4B). Interestingly,CD154+cytokine+ memory T cells reacting to measles were be-low the level of reliable detection in blood in five out of sixdonors, that is, at frequencies below 10−4 of memory CD4+ Tcells and less than 3 × 105 in the entire organ (Fig. 4 A and B andFig. S7). However, in bone marrow, measles-specific CD154+

cytokine+ cells were readily detectable in all donors, at frequencies

BA C

CD

69

Ki67

5.3 0.1

9.6

59.5 0.7

1.8

8.8 0.6

8.1

84.8 0.5

1.8

Memory CD4+ T cells Memory CD8+ T cells

BMPB BMPB

PB BM0

5

10

15 p = 0.0156

MemoryCD4+ T cells

MemoryCD8+ T cells

PB BM PB BM0.0

0.51.010203040 Ex vivo

anti-CD3/28

MemoryCD4+T cells

MemoryCD8+ T cells

Cel

ls in

S+G

2/M

-pha

se (%

)

PB BM0

5

10

15 p = 0.0156

Ki6

7+ce

lls (%

)

Fig. 2. Cell-cycle status of ex vivo human memory CD4+ and CD8+ T cells from BM and PB. (A) Frequencies of Ki67+ memory CD4+ and CD8+ T cells in sevenpaired BM and PB samples. P values were obtained as described in SI Materials and Methods. (B) Sample flow cytometry expression profiles showing thecounterstainings for Ki67 and CD69 expression on memory CD4+ and CD8+ T cells. (C) Frequencies of ex vivo (n = 6) and 72-h anti-CD3/CD28 activated (n = 2)memory CD4+ and CD8+ T cells in the cell-cycle S+G2/M phase. Error bars represent SEM.

Okhrimenko et al. PNAS | June 24, 2014 | vol. 111 | no. 25 | 9231

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 28

, 202

0

of 10−4 to 10−3 of memory CD4+ T cells (Fig. 4A and Fig. S7),corresponding to absolute numbers of 8 × 105 to 1 × 107 cells (Fig.4B). For rubella, CD154+cytokine+ cells were detectable in bonemarrow in similar frequencies as TT-reactive cells, whereas theywere not detectable in blood in two out of four donors (Fig. 4A andFig. S7). However, in the two donors with detectable respondingcells in blood, the frequencies as well as absolute numbers of ru-bella-specific cells were higher in bone marrow than in blood (Fig.4 A and B and Fig. S7). Also similar to TT, mumps-specificCD154+cytokine+ cells were present in higher frequencies in bonemarrow in three out of four donors (Fig. 4 A and B and Fig. S7).The higher frequencies of CD4+ memory T cells in bone marrowrecognizing CMV, TT, measles, rubella, and mumps, comparedwith blood, reflect a preferential recruitment of memory T cellsspecific for systemic antigens to the bone marrow. The exclusivepresence of memory Th cells specific for the measles virus antigenin bone marrow in five out of six donors as well as for the rubellaantigen in two out of four donors shows that the bone marrow isable to maintain long-term memory, even in the apparent absenceof circulating memory cells. For C. albicans-MP65 or vaccinia virus,CD154+cytokine+ memory Th cells were detectable both in bonemarrow and blood in three or four out of five donors, respectively,without significant differences in absolute numbers between thesetwo tissue compartments (Fig. 4 A and B and Fig. S7). The dif-ferential distribution of long-term T-cell memory for systemicpathogens (measles, rubella, and mumps) versus that for skin/mucosal pathogens (C. albican-MP65 and vaccinia) may reflect thedifferent immunization routes and/or different immune responsesas, at least in mice, a prominent skin-resident T-cell memory forvaccinia is known to develop (30).Memory CD4+ T cells that reacted with the reexpression of

two or more of the cytokines TNF-α, IL-2, and IFN-γ are poly-functional and considered to be more protective in secondaryimmune reactions to infectious pathogens than those expressingonly one of the three cytokines (5, 31). Both in terms of fre-quencies and absolute numbers, polyfunctional memory Th cellsspecific for pp65 or TT predominantly resided in bone marrow

(Figs. S7 and S8). In absolute numbers, and with considerableindividual variation, 5–50 times more pp65- or TT-specific pol-yfunctional memory Th cells resided in the bone marrow than inthe blood (Fig. S8). Polyfunctional memory Th cells specific forvaccinia, at frequencies of 0.017–0.189%, corresponding to ab-solute numbers of 4 × 105 to 1.8 × 107, were found both in thebone marrow and in blood (Figs. S7 and S8). Polyfunctionalmemory Th cells specific for measles were only found in the bonemarrow in five out of six donors, at an average frequency of0.016% of memory CD4+ T cells (Figs. S7 and S8).

DiscussionIn the present study, we have compared phenotype, gene expres-sion, repertoire of antigen specificity, and persistence of memoryT cells from bone marrow and blood of individual human donors.Both CD4+ and CD8+ memory T cells from bone marrow arequiescent in their activation status and rest in the G0 phase of thecell cycle, and their gene expression profile suggests that they areimmobile and protected from apoptosis.It has been known for some time that bone marrow contains

a prominent population of CD4+ and CD8+ cells with a memory/effector phenotype, both in humans (32–34) and in mice (1, 35–37). We have shown previously that in murine immune responsesto defined antigens, that are, those resembling vaccines, antigen-specific memory CD4+ T cells translocate to the bone marrowwithin 30–60 d, and thereafter are maintained almost exclusivelyin the bone marrow whereas they slowly decline in blood andSLOs (1). Similarly, the slow contraction of memory T cells inblood has also been described for humans (2), provoking thequestion of whether there is immunological memory withoutmemory T cells (38). Despite this, the analysis of memory T cellsfrom blood has until now continued to be the major basis for theassessment of T-cell memory in humans. However, in light of theapparent contraction of memory T cells in the periphery, mem-ory T cells specific for some pathogens at best represent “short-term” memory. The analysis of murine memory Th cells frombone marrow has put bone marrow memory T cells center-stage,constituting a population of persisting professional memory cellsresulting in stable systemic Th memory (1, 16). Moreover, al-though recently TRM cells have been described in humans (39)and mice (6–8) and are discussed as a major effector of locallong-term memory, it remains unclear as to how they are stablymaintained over time (9).In the present analysis, we also have compared the repertoire

of memory Th cells from human blood and bone marrow (of thesame donors) for specificities representing presumably persistentantigens, namely pp65 of CMV and MP65 of C. albicans; anti-gens encountered within the last 10 y, namely TT; and antigenspresumably encountered only in childhood (40–70 y ago in thecase of our study), namely measles, rubella, mumps, and vacciniavirus or vaccine. In the bone marrow of the donors analyzed,significant populations of memory Th cells specific for any ofthese antigens are present, ranging in numbers from 106 to 108.Memory CD4+ T cells specific for vaccinia were found both in

blood and bone marrow in four out of five analyzed donors,without significant differences in absolute numbers betweenthese two tissue compartments. This could be due to a prefer-ential location of vaccinia-specific memory T cells in the skin, ashas been recently reported in mice (30), and/or a preferentialgeneration of circulating memory cells, compared with bonemarrow memory cells. It has been shown previously that thenumbers of vaccinia-specific CD4+ memory T cells in humanblood decline over time, with a half-life of 8–15 y (2). Also forCandida, CD4+ memory T cells were present in blood at higherfrequencies than in bone marrow, in three out of four donors,suggesting that the repertoire of memory T cells specific forpathogens addressing skin and mucosa are excluded from the

<-S -S/2 0 +S/2 >+SP

B C

D69

- res

t

PB

CD

69- r

est

PB

CD

69- r

est

PB

CD

69- r

est

PB

CD

69- r

est

PB

CD

69- r

est

PB

CD

69- r

est

PB

CD

69- r

est

PB

CD

69-

PB

CD

69+

PB

CD

69-

PB

CD

69-

BM

CD

69+

BM

CD

69+

BM

CD

69-

BM

CD

69-

BM

CD

69-

BM

CD

69-

BM

CD

69-

BM

CD

69+

PB

CD

69- u

nst

PB

CD

69- u

nst

PB

CD

69- u

nst

PB

CD

69- u

nst

PB

CD

69- u

nst

PB

CD

69- u

nst

PB

CD

69- u

nst

PB

CD

69- u

nst

Fig. 3. Global resting gene expression profiles of ex vivo memory CD4+ Tcells from BM and PB. Transcriptomes of memory CD4+ T-cell subsets fromfour paired BM and PB samples were compared with transcriptomes of CD4+

memory T cells from PB of eight unrelated donors, directly after isolation(ex vivo) or after stimulation for 3 h with PMA/ionomycin. Differentiallyexpressed genes (7,383) were hierarchically clustered and displayed. Thenormalized induction (red) or repression (blue) is shown for each gene.Transcriptomes of the following samples are shown: from paired ex vivo BMand PB samples, BM CD69+, BM CD69−, and PB CD69−; from PB samples ofunrelated donors, PB CD69− unstimulated (unst) or restimulated with PMA/ionomycin (rest).

9232 | www.pnas.org/cgi/doi/10.1073/pnas.1318731111 Okhrimenko et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 28

, 202

0

bone marrow and may be maintained in the tissues affected astissue-resident memory cells (30, 40).CD4+ memory Th cells specific for measles were exclusively

detectable in bone marrow in five out of six donors, in numberscomparable to TT-specific cells. Also, in two out of four donors,CD4+ memory T cells specific for rubella were only detectable inbone marrow, and not in blood. Memory Th cells specific forCMV, TT, mumps, and rubella, the latter in two out of fourdonors, were also found in the blood, but at frequencies amongCD4+ memory T cells of at least twofold less than in bonemarrow. Thus, the repertoire of human bone marrow CD4+

T-cell memory is significantly enriched for T cells specific forsystemic pathogens and vaccines. In terms of absolute cellnumbers, CD4+ T cells specific for these systemic antigens wereeven more prominent, 5–20 times, in the bone marrow than inthe blood. This observation is in line with a stable maintenanceof memory Th cells in the bone marrow, whereas their numberswould dwindle in the blood (1). Moreover, this observationshows that memory Th cells in the bone marrow can be main-tained in the apparent absence of memory Th cells in blood, inother words, that the bone marrow cells are not circulating butremain residents of the bone marrow.This notion of bone marrow residency is supported by differ-

ences in the phenotypes of memory T cells from blood and bonemarrow. For example, we found that 30% of CD4+ and 60%of CD8+ memory T cells from bone marrow expressed CD69,commonly regarded as an activation marker (21). Accordingly,memory T cells from bone marrow have been considered to be ina “heightened” activation state (10, 41). However, here we showthat despite expression of CD69, these cells are quiescent. Theydo not express other activation markers, such as CD137, theyshow a gene transcription profile of resting cells, very similar tothat of memory T cells from blood and clearly distinct from thatof activated T cells, and they are in the G0 phase of the cell cycle,as determined by Ki67 staining. This resembles the situation inthe mouse, where 30% of the memory Th cells of bone marrowexpress CD69 yet are resting in terms of proliferation and overall

transcriptional activity (1). In mice, expression of CD69 on ac-tivated Th cells is critical for their migration into the bonemarrow and the establishment of Th cell memory as such (17).Furthermore, expression of CD69 may also be critical to retainthe cells in the bone marrow, because expression of a receptorfor S1P, which mediates egress into the blood (25–27), is down-regulated in CD69+ bone marrow memory Th cells.Bone marrow CD4+ memory T cells are imprinted for efficient

reexpression of effector cytokines. Memory cells specific forpp65 of CMV, TT, measles, rubella, mumps, MP65 of C. albi-cans, or vaccinia, when restimulated, readily expressed two orthree of the cytokines TNF-α, IL-2, and IFN-γ, marking them as“polyfunctional” memory T cells. In immune responses to HIV,it had been shown that this polyfunctionality is essential for ef-ficient immune responses to the pathogen (5, 31). For systemicpathogens (CMV-pp65, TT, measles, rubella, and mumps), specificpolyfunctional memory T cells preferentially reside in bone marrow.Even if one considers that not all cycling memory T cells are in theblood but more are in the SLOs (42) and tissues such as the skin(43), CMV-, TT-, and measles-specific polyfunctional memoryT cells are largely restricted to the bone marrow. Interestingly,polyfunctional CD4+ memory T cells specific for C. albicans-MP65- and vaccinia-specific ones are not, again indicating thatCD4 T-cell memory for these antigens may be maintained in thetissues affected, namely skin and mucosa.From the transcriptomes of bone marrow memory cells, it is

obvious that all of the lineage-determining transcription factorsof memory T-cell differentiation are expressed, including TBX21,GATA3, RORC, EOMES, and BCL6. FOXP3 is also expressed,suggesting that a fraction of the memory cells in bone marroware regulatory memory T cells. Indeed, 2.1% of the CD69+ and5.5% of the CD69− Th memory cells of bone marrow have thephenotype CD25+CD127low and FOXP3+, qualifying them asbona fide Treg cells.As we show here, memory T cells of the bone marrow are

resting and maintain antigen-specific memory over long periodsof time, presumably decades, even when memory T cells of the

CMV-pp65 TT Measles Vaccinia

PB BM0.0

0.5

1.0

1.5 p = 0.0156

CD

154+ cy

toki

ne+

cells

(%)

PB BM0.00

0.04

0.08

0.12 p = 0.3125

PB BM0.0

0.1

0.2

0.3 p = 0.4375

PB BM0.00

0.04

0.08

0.12 p = 0.0313

Rubella Mumps C. albicans-MP65

PB BM0.0

0.1

0.2

0.3 p = 0.125

PB BM0.00

0.05

0.10

0.15 p = 0.375

PB BM0.00

0.02

0.04

0.06

0.08 p = 0.4375

A

CMV-pp65 TT Measles Vaccinia

PB BM0

0.2

0.4

0.6

0.8 p = 0.0625

PB BM0

5

10

15 p = 0.0156

CD

154+ cy

toki

ne+

cells

(x10

7 )

PB BM0

1

2

3p = 0.1875

PB BM0

0.5

1.0

1.5 p = 0.0313

Rubella Mumps C. albicans-MP65

PB BM0

0.5

1.0

1.5

2.0 p = 0.125

PB BM0

0.5

1.0

1.5 p = 0.25

PB BM0

0.1

0.2

0.3

0.4

0.5 p = 0.4375

B

Fig. 4. Comparison of antigen specificity and diversity of memory CD4+ T cells from paired BM and PB samples. Mononuclear cells isolated from four to sevenpaired BM and PB samples were stimulated with the indicated antigens, and the induced cytokine production (IFN-γ, IL-2, or TNF-α) in memory CD4+ T cellswas examined according to CD154 expression. For each subpopulation, the background (as detected in the anti-CD28 stimulated but otherwise equallytreated control samples) was subtracted. (A) Frequencies and (B) estimated absolute numbers of antigen-specific CD154+cytokine+ (total cytokine-producing)cells are shown. Absolute numbers of antigen-specific memory CD4+ T cells were calculated using the frequency of these cells and the estimated numbers ofT cells within BM and PB of healthy young adults (18, 19). Symbols in gray indicate frequencies under reliable detection limit (10−4 of memory CD4+ T cells) orestimated cell numbers calculated according to such frequencies. P values were obtained as described in SI Materials and Methods.

Okhrimenko et al. PNAS | June 24, 2014 | vol. 111 | no. 25 | 9233

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 28

, 202

0

same specificity are no longer detectable in the blood. This isa drastic reversal of the current concept that memory T cellsrecirculate throughout the body to find their cognate antigen,presented to them by APCs in SLOs. If memory T cells resideand rest in the bone marrow, antigen-sampling and/or -presentingcells would have to come to the bone marrow to find and mobilizethe memory T cells specific for their antigen. At present itremains enigmatic how this works. However, it has been shownalready for murine memory Th cells from bone marrow that theyare superior helper cells in secondary immune responses, com-pared with memory T cells from spleen, efficiently promotingantibody class switching and affinity maturation of activated Bcells (1, 17).

Materials and MethodsSample Collection and Preparation. Unpaired blood samples were collectedfrom anonymous healthy adult donors (mean age ± SEM: 47.5 ± 3.6; n = 8)and paired bone marrow and peripheral blood samples from anonymoussystemically healthy adults (63.8 ± 1.2; n = 61) undergoing hip replacementoperations. All samples were obtained with local ethical committee(Ethikkommission der Charité-Univerisitätsmedizin Berlin) approvals andinformed consent in accordance with the Declaration of Helsinki. All freshlyobtained samples were subjected to immediate preparation. Mononuclearcells of blood and bone marrow were isolated by density gradient sedi-mentation using Ficoll-Hypaque (Sigma-Aldrich).

Flow Cytometry Analysis. Eight- to 12-color flow cytometry was performed forthe analysis of cell phenotype, cytokine profile, and cell sorting using a BDFACSAria cell sorter, LSRFortessa flow cytometer (BD Biosciences), orMACSQuant (Miltenyi Biotec). The following fluorochrome-conjugatedmouse

anti-human antibodies were used to stain cells: CD14 (Beckman Coulter), CD19and CD4 (house conjugate), Ki67 (eBioscience), CD154 (Miltenyi Biotec), CD3,CD45RA, and CD45RO (BD Biosciences), CD8, CD69, CD25, CD127, CCR7, CD137,TNF-α, IFN-γ, and IL-2 (BioLegend). DAPI or PI was used as a dead cell exclusionmarker. Stained cells were acquired using FACSDiva (BD Biosciences) or MACS-Quantify software and data were analyzed with FlowJo (Tree Star).

FOXP3 Staining. Cells isolated from paired bone marrow and blood sampleswere first stained for surface markers followed by fixation/permeabilizationand intracellular staining using the Foxp3 Staining Buffer Set (eBioscience)according to the manufacturer’s recommendations.

Cell-Cycle Analysis. Cell-cycle status was assessed by staining cells with Ki67 orPI. Briefly, live cells were stained with surface markers followed by fixation/permeabilization using FACS Lysing Solution/FACS Permeabilizing Solution 2(BD Biosciences) for Ki67 or Foxp3 Fix/Perm buffers (eBioscience) for PI. In thecase of PI staining, cells were treated with RNaseA (200 μg/mL) and stainedfor PI (20 μg/mL) in PBS at 37 °C for 15 min. Samples were acquired viaMACSQuant (Miltenyi Biotec).

Additional methods are available in SI Materials and Methods.

ACKNOWLEDGMENTS. We are thankful to Antje Blankenstein and the CoreUnit Cell Harvesting at the Berlin-Brandenburg Center for RegenerativeTherapies for providing paired peripheral blood and bone marrow samples.We acknowledge the assistance of the Flow Cytometry Core Facility at DeutschesRheuma-Forschungszentrum Berlin. R.K. and C.S. were in part supported by TheJosé Carreras Leukaemia Foundation. This work was supported by a EuropeanResearch Council (ERC) Advanced Grant [ERC-2010-AdG_20100317 Grant 268987(to A.R.)] and Priority Programme 1468 Immunobone of the German ResearchCouncil, and in part by the International Max Planck Research School for Infec-tious Diseases and Immunology Generation 2011.

1. Tokoyoda K, et al. (2009) Professional memory CD4+ T lymphocytes preferentiallyreside and rest in the bone marrow. Immunity 30(5):721–730.

2. Hammarlund E, et al. (2003) Duration of antiviral immunity after smallpox vaccina-tion. Nat Med 9(9):1131–1137.

3. Sallusto F, Geginat J, Lanzavecchia A (2004) Central memory and effector memory Tcell subsets: Function, generation, and maintenance. Annu Rev Immunol 22:745–763.

4. Löhning M, Richter A, Radbruch A (2002) Cytokine memory of T helper lymphocytes.Adv Immunol 80:115–181.

5. Betts MR, et al. (2006) HIV nonprogressors preferentially maintain highly functionalHIV-specific CD8+ T cells. Blood 107(12):4781–4789.

6. Masopust D, Vezys V, Marzo AL, Lefrançois L (2001) Preferential localization of ef-fector memory cells in nonlymphoid tissue. Science 291(5512):2413–2417.

7. Gebhardt T, et al. (2009) Memory T cells in nonlymphoid tissue that provide enhancedlocal immunity during infection with herpes simplex virus. Nat Immunol 10(5):524–530.

8. HofmannM, Oschowitzer A, Kurzhals SR, Krüger CC, Pircher H (2013) Thymus-residentmemory CD8+ T cells mediate local immunity. Eur J Immunol 43(9):2295–2304.

9. Bevan MJ (2011) Memory T cells as an occupying force. Eur J Immunol 41(5):1192–1195.10. Herndler-Brandstetter D, et al. (2011) Human bone marrow hosts polyfunctional

memory CD4+ and CD8+ T cells with close contact to IL-15-producing cells. J Immunol186(12):6965–6971.

11. Slifka MK, Matloubian M, Ahmed R (1995) Bone marrow is a major site of long-termantibody production after acute viral infection. J Virol 69(3):1895–1902.

12. Manz RA, Thiel A, Radbruch A (1997) Lifetime of plasma cells in the bone marrow.Nature 388(6638):133–134.

13. Tokoyoda K, Egawa T, Sugiyama T, Choi BI, Nagasawa T (2004) Cellular niches con-trolling B lymphocyte behavior within bone marrow during development. Immunity20(6):707–718.

14. Kondrack RM, et al. (2003) Interleukin 7 regulates the survival and generation ofmemory CD4 cells. J Exp Med 198(12):1797–1806.

15. Li J, Huston G, Swain SL (2003) IL-7 promotes the transition of CD4 effectors to per-sistent memory cells. J Exp Med 198(12):1807–1815.

16. Tokoyoda K, Hauser AE, Nakayama T, Radbruch A (2010) Organization of immuno-logical memory by bone marrow stroma. Nat Rev Immunol 10(3):193–200.

17. Shinoda K, et al. (2012) Type II membrane protein CD69 regulates the formation ofresting T-helper memory. Proc Natl Acad Sci USA 109(19):7409–7414.

18. Trepel F (1974) Number and distribution of lymphocytes in man. A critical analysis.Klin Wochenschr 52(11):511–515.

19. Di Rosa F, Pabst R (2005) The bone marrow: A nest for migratory memory T cells.Trends Immunol 26(7):360–366.

20. Ziegler SF, et al. (1993) Molecular characterization of the early activation antigenCD69: A type II membrane glycoprotein related to a family of natural killer cell ac-tivation antigens. Eur J Immunol 23(7):1643–1648.

21. Testi R, Phillips JH, Lanier LL (1989) T cell activation via Leu-23 (CD69). J Immunol143(4):1123–1128.

22. Miyara M, et al. (2009) Functional delineation and differentiation dynamics of humanCD4+ T cells expressing the FoxP3 transcription factor. Immunity 30(6):899–911.

23. Gerdes J, Schwab U, Lemke H, Stein H (1983) Production of a mouse monoclonal

antibody reactive with a human nuclear antigen associated with cell proliferation. Int

J Cancer 31(1):13–20.24. Paiardini M, et al. (2005) Loss of CD127 expression defines an expansion of effector

CD8+ T cells in HIV-infected individuals. J Immunol 174(5):2900–2909.25. Shiow LR, et al. (2006) CD69 acts downstream of interferon-alpha/beta to inhibit S1P1

and lymphocyte egress from lymphoid organs. Nature 440(7083):540–544.26. Matloubian M, et al. (2004) Lymphocyte egress from thymus and peripheral lymphoid

organs is dependent on S1P receptor 1. Nature 427(6972):355–360.27. Feng C, et al. (2002) A potential role for CD69 in thymocyte emigration. Int Immunol

14(6):535–544.28. Frentsch M, et al. (2005) Direct access to CD4+ T cells specific for defined antigens

according to CD154 expression. Nat Med 11(10):1118–1124.29. Chattopadhyay PK, Yu J, Roederer M (2005) A live-cell assay to detect antigen-specific

CD4+ T cells with diverse cytokine profiles. Nat Med 11(10):1113–1117.30. Jiang X, et al. (2012) Skin infection generates non-migratory memory CD8+ T(RM) cells

providing global skin immunity. Nature 483(7388):227–231.31. Seder RA, Darrah PA, Roederer M (2008) T-cell quality in memory and protection:

Implications for vaccine design. Nat Rev Immunol 8(4):247–258.32. Feuerer M, et al. (2001) Therapy of human tumors in NOD/SCID mice with patient-

derived reactivated memory T cells from bone marrow. Nat Med 7(4):452–458.33. Palendira U, et al. (2008) Selective accumulation of virus-specific CD8+ T cells with

unique homing phenotype within the human bone marrow. Blood 112(8):3293–3302.34. Guerreiro M, et al. (2010) Human peripheral blood and bone marrow Epstein-Barr

virus-specific T-cell repertoire in latent infection reveals distinct memory T-cell sub-

sets. Eur J Immunol 40(6):1566–1576.35. Becker TC, Coley SM, Wherry EJ, Ahmed R (2005) Bone marrow is a preferred site for

homeostatic proliferation of memory CD8 T cells. J Immunol 174(3):1269–1273.36. Mazo IB, et al. (2005) Bone marrow is a major reservoir and site of recruitment for

central memory CD8+ T cells. Immunity 22(2):259–270.37. Quinci AC, et al. (2012) IL-15 inhibits IL-7Rα expression by memory-phenotype CD8+ T

cells in the bone marrow. Eur J Immunol 42(5):1129–1139.38. Bell EB, Westermann J (2008) CD4 memory T cells on trial: Immunological memory

without a memory T cell. Trends Immunol 29(9):405–411.39. Sathaliyawala T, et al. (2013) Distribution and compartmentalization of human cir-

culating and tissue-resident memory T cell subsets. Immunity 38(1):187–197.40. Masopust D, et al. (2010) Dynamic T cell migration program provides resident memory

within intestinal epithelium. J Exp Med 207(3):553–564.41. Price PW, Cerny J (1999) Characterization of CD4+ T cells in mouse bone marrow. I.

Increased activated/memory phenotype and altered TCR Vbeta repertoire. Eur J Im-

munol 29(3):1051–1056.42. Klonowski KD, et al. (2004) Dynamics of blood-borne CD8 memory T cell migration in

vivo. Immunity 20(5):551–562.43. Clark RA, et al. (2006) The vast majority of CLA+ T cells are resident in normal skin.

J Immunol 176(7):4431–4439.

9234 | www.pnas.org/cgi/doi/10.1073/pnas.1318731111 Okhrimenko et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 28

, 202

0