Tissue-resident memory CD8+ T cells continuouslypatrol skin epithelia to quickly recognize local antigenSilvia Ariottia, Joost B. Beltmanb,1, Grzegorz Chodaczekc,1, Mirjam E. Hoekstraa, Anna E. van Beeka,Raquel Gomez-Eerlanda, Laila Ritsmad, Jacco van Rheenend, Athanasius F. M. Maréee, Tomasz Zalc,Rob J. de Boerb, John B. A. G. Haanena, and Ton N. Schumachera,2

aDivision of Immunology, The Netherlands Cancer Institute, 1066 CX, Amsterdam, The Netherlands; bTheoretical Biology and Bioinformatics, UtrechtUniversity, 3584 CH, Utrecht, The Netherlands; cDepartment of Immunology, University of Texas, MD Anderson Cancer Center, Houston, TX 77054; dHubrechtInstitute-Royal Netherlands Academy of Arts and Sciences and University Medical Center Utrecht, 3584 CT, Utrecht, The Netherlands; and eComputational andSystem Biology, John Innes Centre, Norwich NR4 7UH, United Kingdom

Edited by Rafi Ahmed, Emory University, Atlanta, GA, and approved October 9, 2012 (received for review May 30, 2012)

Recent work has demonstrated that following the clearance ofinfection a stable population of memory T cells remains present inperipheral organs and contributes to the control of secondaryinfections. However, little is known about how tissue-residentmemory T cells behave in situ and how they encounter newlyinfected target cells. Here we demonstrate that antigen-specificCD8+ T cells that remain in skin following herpes simplex virus infec-tion show a steady-state crawling behavior in between keratino-cytes. Spatially explicit simulations of the migration of thesetissue-resident memory T cells indicate that the migratory dendriticbehavior of these cells allows the detection of antigen-expressingtarget cells in physiologically relevant time frames of minutes tohours. Furthermore, we provide direct evidence for the identifica-tion of rare antigen-expressing epithelial cells by skin-patrollingmemory T cells in vivo. These data demonstrate the existence ofskin patrol by memory T cells and reveal the value of this patrolin the rapid detection of renewed infections at a previouslyinfected site.

Cellular Potts Model | intravital imaging | HSV-1

After the clearance of infection, the majority of antigen-specific T cells that have expanded to contribute to pathogen

control die, but a small fraction of memory T cells (Tmem) surviveto confer protection upon reencountering the same pathogen.Most of these protective cells continuously circulate betweenblood and lymph nodes; however, some peripheral cells, so-calledtissue-resident memory T cells, permanently abandon the circu-lating memory pool and take up residence in nonlymphoid tissues(1, 2). Tissue-resident memory T cells have been observed insensory ganglia (3), brain parenchyma (4), and both the intestinal(5) and skin epithelia (2, 3). At least in the case of skin epithelia,resident memory T cells exist in disequilibrium with the circulatingpool and can therefore be considered as a distinctive memorysubset (2, 6, 7). In the absence of sustained antigen presentation(i.e., following virus clearance), skin-resident memory T cellsdown-regulate transcription of cytolytic molecules (8). Neverthe-less, tissue-resident memory T cells provide superior protectionagainst local virus reactivation relative to the circulating memoryT-cell pool (3, 6, 9).Although it has been proposed that the control of local infec-

tions by resident memory T cells is due to the advantage conferredby the proximity to sites of viral entry (6), it has not been resolvedhow tissue-residing memory T cells mediate their protective effect.Here we address this question, through a combination of intravitalmicroscopy and computational modeling. We demonstrate thatthe CD8+ T cells that stay behind in the epidermis actively patrolthe skin, continuously migrating within the host tissue to contactsurrounding cells. This uninterrupted movement, combined withthe dendritic shape that memory T cells acquire upon epithelialresidence, enables them to encounter infected cells in physiologi-cally relevant timeframes. Thus, morphology and motility of tissue-

resident memory T cells constitute a central component of theirability to provide enhanced local immune protection.

ResultsModel to Dissect the Function of Skin-Resident Memory T Cells. Tounderstand how tissue-resident CD8+ memory T cells guardagainst viral reinfections, we first set out to develop an experi-mental model in which the activity of skin-resident versus systemicCD8+ Tmem can be analyzed without potential confoundingeffects of virus-specific CD4+memory T cells or antibodies. To thispurpose, we transferred naive CD8+ herpes simplex virus-1(HSV)–specific (10), GFP+ T cells (GFP+gB) into C57BL/6jrecipients and subsequently vaccinated the mice by skin DNAinjection, using a DNA vaccine encoding the gB498–505 epitope thatis recognized by gB T cells (Fig. 1A). The magnitude of vaccine-induced gB498–505-specific CD8+ T-cell responses in these mice (upto 5% of circulating CD8+ T cells at days 11–15 postvaccination;Fig. 1B) was comparable to that induced in mice that had notreceived adoptive transfer. However, the gB498–505-specific CD8+

T-cell response in recipients of GFP+ gB T cells fully consisted ofGFP+ cells (Fig. S1A), thus ensuring that the fluorescent cells ob-served during imaging represented the entire population of gB498–

505-responsive T cells. Analysis of the vaccination site several weeksafter treatment revealed the presence of large numbers of GFP+

cells that displayed a dendritic morphology (Fig. 1C). In agreementwith previous literature (2), multiphoton analysis of the skinestablished that the GFP+ cells were exclusively restricted to theepidermal layers, including hair shafts (Fig. 1 D and E). Flowcytometry demonstrated that the GFP+ cells extracted from theskin were uniformly CD8+TCRβ+ and KbgB498–505tetramer+,thereby qualifying them as bona fide HSV-specific CD8+ Tmem(Fig. 1F); furthermore, they expressed a high level of CD103(integrin αEβ7; Fig. 1G), an adhesion molecule implicated in T-cellretention within the epithelia (6). Consistent with published liter-ature (6), the skin-resident memory T-cell pool that was induced byskin vaccination displayed a morphology and phenotype similar tothat found after local skin infection with HSV, demonstrating thatlong-term tissue retention of CD8+ Tmem is not dependent on thepresence of virus. To determine the contribution of the vaccine-induced CD8+ skin-resident memory T-cell pool to the control of

Author contributions: S.A., J.B.B., and T.N.S. designed research; S.A., J.B.B., G.C., and L.R.performed research; M.E.H., A.E.v.B., R.G.-E., and A.F.M.M. contributed new reagents/analytic tools; S.A., J.B.B., J.v.R., T.Z., and J.B.A.G.H. analyzed data; and S.A., J.B.B.,R.J.d.B., and T.N.S. 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.1J.B.B. and G.C. contributed equally to this work.2To whom correspondence should be addressed. E-mail: t.schumacher@nki.nl.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1208927109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1208927109 PNAS | November 27, 2012 | vol. 109 | no. 48 | 19739–19744

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

secondary infections, we vaccinated recipients of GFP+gB T cellson one flank with the gB498–505 DNA vaccine, to create a localizedCD8+ memory T-cell pool. Three months after vaccination, micewere challenged with an HSV strain that expresses the td-tomatofluorescent protein (HSVTOM) at both the ipsilateral (i.e., pre-viously vaccinated) and contralateral site, and infection growth wasmonitored over time (Fig. S1B). Intravital imaging of the infectedskin revealed the presence of similarly sized HSV-infected lesionsearly after infection (P = 0.4402). In contrast, whereas infections atthe contralateral site continued to grow rapidly, HSV foci atmemory CD8+ T-cell sites were in large part contained (Fig. S1 Cand D; comparison contra – ipsi, P < 0.001), establishing that skin-resident CD8+ T cells play an important role in the control of localsecondary HSV infections.

Skin-Resident Memory T Cells Display a Continuous Migratory Behavior.To investigate how tissue-resident memory T cells can provide

such local protection we performed long-term imaging of skin-resident GFP+CD8+ Tmem at formerly infected sites. Theseexperiments revealed that at all of the examined time pointspostinfection, these cells were not sessile, but migrated pro-ductively within the epidermis, with mean speeds of about1.3 μm/min (Movie S1). Whereas the speed of Tmem migrationwas relatively low, Tmem migration was characterized by longdirectional persistence, with cells moving in a similar directionfor an average of ∼10.5 min. Combined, these speed and di-rectionality parameters result in a high motility coefficient of∼9 μm2/min, close to that of lymph node B cells (11). To addresswhether this steady-state migration could reflect continued low-level T-cell receptor (TCR) triggering due to the presence ofresidual antigen, we generated an experimental setting in whichtissue-resident memory T cells reside in skin areas in which theircognate antigen was never introduced. To this purpose, naiveGFP+OTI T cells were transferred into C57BL/6j recipientsand activated by application of a DNA vaccine encoding theOVA257–264 epitope that is recognized by the OTI TCR to createa pool of effector GFP+OTI T cells. At this time point, micewere infected by epidermal HSV application, and entry andmaintenance of the “bystander” GFP+OTI T cells at the site ofHSV infection was evaluated. These analyses demonstrated thateffector CD8+ T cells also form long-term local T-cell memory atsites of former bystander infection (Fig. 2A). Importantly, thesebystander tissue-resident memory T cells displayed a dendriticmorphology (Fig. 2B) and showed a productive migrationthrough the epidermis (Movie S2), indicating that both of theseTmem cell properties are antigen-independent. Thus, regardlessof the presence of cognate antigen, CD8+ T cells that enter theskin epithelium up-regulate the surface expression of CD103 andacquire dendritic morphology and constitutive motility.

Migratory Behavior of Memory T Cells Allows Contact with a LargeNumber of Surrounding Cells. The continuous movement of theepithelium-resident memory pool (Fig. 2C) suggests that in timeCD8+ T cells can cover long distances and as a consequencecontact a large number of surrounding tissue cells. To visualizethe cumulative skin area that can be patrolled by migratingpopulations of skin-resident CD8+ Tmem over time, all con-secutive images from a 4-h imaging session were overlayed (Fig.2D). This analysis revealed that at the cell densities found withinpreviously infected areas, the migratory behavior of CD8+ Tmemis sufficient to allow the exploration of a large portion of skinwithin a period of hours. Constitutive steady-state migration ofskin-resident Tmem was observed up to 18 mo after primaryinfection, the latest time point analyzed. As a control, and con-sistent with prior data (12), the cell body of skin-residentLangerhans cells (LCs) did not show detectable displacement(Fig. 2E). LCs, epidermal γδ T cells, and CD8+ Tmem within theepidermis all have a dendritic shape. It has been established thatLCs project their dendrites upward, toward the stratum corneum(13), a property that allows them to capture antigens from themost superficial skin layers. Likewise, the dendrites of γδ T cellsinvariably project toward the stratum corneum (14). However,the observation that skin-resident CD8+ Tmem show a continu-ous migration through the epidermis, together with the fact thatthe dendrites continuously change size, length, and direction(Movies S1 and S2), suggests that these dendrites could servea different function. To compare the morphology of dendritesfrom skin-resident γδ T cells, LC, and CD8+ Tmem we analyzedthese cell types in whole-mount stains. CD8+ Tmem had fewerbut thicker dendrites compared with γδ T cells; more impor-tantly, whereas the dendritic extensions of Langerhans cells andepidermal γδ T cells pointed upward, the dendrites of CD8+

Tmem were almost all parallel to the epidermis, consistent withpatrol along a 2D surface (Fig. 3 and Movie S3). Thus, memoryT cells residing in the epidermis differ from other immune cells

BA

C

D

0

0.4

0.8

1.2

0 20 40 60 80 100

Pix

el in

tens

ity(n

orm

aliz

ed)

Whole field analysis

Z-depth (µm)

post 1st infection - Tmem

T cells - Type I Collagen (Dermis)

GFP

TCRbeta CD8

gB-te

tram

er

CD103

coun

ts

E

F

- 1 0 3 6 >35

GFP+gB transfer

gB vaccination

imaging

l---epi---l l------dermis------l

T cells - Type I Collagen (Dermis)

0 10 200

2

4

6

180

days post adoptive transfer

% o

fCD

8+ce

lls

G

Fig. 1. Morphology, localization, and phenotype of skin-resident memoryCD8+ T cells. (A) Recipients of GFP+gB T cells (day −1) were vaccinated witha DNA vaccine encoding the gB498–505 epitope (days 0, 3, and 6) and imaged1 mo later. (B) The percentage of GFP+gB498–505tetramer+ cells was measuredin blood over time (arrows indicate vaccination timepoints). (C) Top viewconfocal maximum intensity projections of GFP+gB T cells infiltrating skin 1mo after local HSVTOM infection. Images are representative of five experi-ments. (Scale bars: Left, 50 μm; Middle and Right, 10 μm.) (D and E) Areas ofskin containing infiltrating GFP+gB T cells were imaged by intravital multi-photon microscopy 3 mo after infection; the position of the dermis wasdetermined by exploiting second harmonic generation properties of colla-gen type I. Arrows indicate tissue direction, from outside to inside the ani-mal. (F) Cell suspensions from skin biopsies 1 mo after local HSVTOM infectionwere analyzed for the indicated markers and GFP expression. Left plot isgated on live cells; right plot is gated on TCRβ+ live cells. GFP+ cells aredepicted in green and GFP− cells are depicted in black. (G) Surface expressionof CD103 by TCRβ+CD8+GFP+ cells from previously infected skin (green), TCRβ+

CD8+GFP+ splenocytes (gray), and TCRβ+CD8+GFP− splenocytes (black line).

19740 | www.pnas.org/cgi/doi/10.1073/pnas.1208927109 Ariotti et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

within the same environment with respect to their morphologyand migration.

Dendritic Shape of Skin-Resident Memory T Cells Is Independent ofSkin Inflammatory State. Effector CD8+ T cells, which are presentin both the epidermis and dermis of HSV-infected mice, exhibit anamoeboid shape, whereas skin-resident memory CD8+ T cells,which are located exclusively in the epidermis (2; Fig. 1 D and E),display a dendritic shape. To evaluate whether the shape differ-ence between the two cell populations could be due to the skinlayer in which they reside, we compared the shape of effectorT cells (Teff) and Tmem in the same skin layer using LCs as aninternal reference (Fig. S2). This analysis demonstrated that theshape difference between the two is not due to a simple differencein location. A second explanation for the “dendricity” acquired by

memory T cells upon retention in the epidermis could be that thisshape is imposed on CD8+ Tmem by the tight surrounding tissueafter infection resolves. To investigate whether the dendritic shapedisplayed by the memory T-cell pool in the epidermis is solelyimposed by the environment, we compared the shape of GFP+

OTI T cells present at sites of ongoing HSV infections, when ei-ther present as recently primed effector T cells or as locally resi-dent memory T cells (Fig. 4A). Effector OTI T cells migratingwithin HSV-infected skin displayed the amoeboid, round shapethat is also seen for antigen-specific effector T cells within effectorsites (Fig. 4B). In contrast, skin-resident memory OTI T cells thatinfiltrated an HSVTOM-infected area retained their dendritic,crawling behavior (Fig. 4 B and C and Movie S4). Thus, thedendritic migration mode of CD8+ Tmem that are present in theepidermis does not merely stem from the physical constraintsimposed by the environmental architecture, but at least partiallyreflects an adaptation of the memory T cells present at that site.The observation that skin-resident CD8+ Tmem that move withinthe epidermis continuously extend and contract dendrites alonga 2D plane suggests that these cells migrate in between kerati-nocytes. To visualize the cellular environment of skin-residentCD8+ Tmem, we transferred naive Kaede+gB T cells (15) intoHistone-2B-GFP mice (in which GFP marks the nuclei of all skincells) and created a local skin-resident CD8+ Tmem pool. Severalweeks postinfection, Kaede was locally photoconverted in vivo, todistinguish the transferred T cells from the GFP+ skin cells. Im-aging of the photoconverted areas confirmed that skin-residentCD8+ Tmem extend and retract dendrites continuously in be-tween skin cells while moving through the epidermis (Fig. 4D andMovie S5). To evaluate whether the differential behavior ofTmem and Teff correlated with phenotypic differences, the sur-face expression of a small number of T-cell markers was studied.CCR4 and NKG2D expression did not differ between Tmem andTeff (Fig. S3), making their contribution to the differential be-havior of the two cell types unlikely. The chemokine receptorCXCR3, involved in T-cell migration to peripheral tissues (16),was detected on skin-resident memory T cells but was nearly ab-sent on HSV-infected skin-infiltrating effector T cells (Fig. S3),conceivably reflecting down-regulation induced by ligand binding.

0h

0h 4h

- 1 0 3 6 35

GFP+OTI

OVA vaccinationleft flank

imaging

HSVright flank

10

A

B

0h

0h 4h

5’ to 10’ 10’ to 15’ 15’ to 20’

20’ to 25’ 25’ to 30’ 30’ to 35’ 35’ to 40’

40’ to 45’ 45’ to 50’ 50’ to 55’ 55’ to 60’

0’ to 5’EDC

Fig. 2. Steady-state, skin-resident memory CD8+ T cells migrate in between epithelial cells. (A and B) Recipients of GFP+OTI T cells (day −1) were vaccinatedwith a DNA vaccine encoding the Ova257–264 epitope (days 0, 3, and 6), infected with HSVTOM (day 10), and imaged at the site of HSVTOM infection 1 mo later.Top view confocal maximum intensity projections of GFP+OTI T cells infiltrating skin 1 mo after local HSVTOM infection are depicted. (C) Migration ofa representative GFP+gB T cell residing in epidermis recovered from a previous HSVTOM infection; each snapshot is the superimposition of five consecutiveimages taken at 1-min intervals. (Scale bar, 10 μm.) (D and E, Upper) First image of a 4-h time-lapse imaging session of GFP+gB T cells (D) or GFP+ LCs cells (E)migrating through skin 1 mo after local HSVTOM infection. (Lower) Superimposition of all images from the 4-h time-lapse imaging session, indicating the areaexplored by GFP+gB T cells and GFP+ LCs cells during that time interval. Images are representative of two experiments. (Scale bar, 20 μm.)

-60

-30

30

60

90p<0.0001

pro

ject

ion

angl

e (d

egre

es)

A

0

θθθ

B

γδ Tcell

Tmem

γδ Tcell

C

Tmem

Fig. 3. Skin-resident memory CD8+ T cells are morphologically differentfrom γδ T cells. (A and B) Skin biopsies containing tissue-resident memory Tcells were stained to detect γδ T cells and memory T cells. Two-dimensionalmasks of the cellular bodies were used to calculate dendrite projectionangles. The angles were measured based on side projections of the singlecells and calculated between two vectors starting from the cell’s center ofmass, one vector being parallel to the stratum corneum, the other goingthrough the dendrite tip, as depicted in A. The results from 16 Tmem (28dendrites) and 18 γδ T cells (92 dendrites) are shown. Bars indicate median.(C) Representative image of a skin-resident memory CD8+ T cell and Lang-erhans cell with dendrites protruding sideways and upward, respectively.

Ariotti et al. PNAS | November 27, 2012 | vol. 109 | no. 48 | 19741

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

Quantification of the Efficiency of Target Cell Encounter byComputational Modeling. The above data demonstrate that mem-ory T cells continuously migrate through the epidermal surface forlong periods after infection and in the absence of antigenic stim-ulation. These observations raise the hypothesis that the combi-nation of migratory behavior and morphology could representa type of tissue surveillance, in which CD8+ Tmem patrol theepidermis to search for skin cells that express cognate antigen. Toexplore to what extent the migration and morphology of Tmemwould increase the contact and scanning of surrounding cells, wedeveloped a 2D Cellular Potts Model (17–20), in which the densityof keratinocytes (21) and T-cell migration and shape parametersclosely match the experimentally determined values (SI Materialsand Methods, Movie S6, Fig. 5A, and Fig. S4 A and B). Sub-sequently, simulated “target cells” and memory T cells were po-sitioned randomly within an in silico field (size 0.54 mm2) andTmem–target encounters were monitored over time. At the ex-perimentally determined Tmem densities of about 430 cells persquare millimeter, the majority of targets was detected within 1 h(Fig. 5B). Furthermore, even at T-cell densities that were up tofivefold below those determined experimentally, most targets wereencountered within hours. If T-cell densities are reduced even

further, the efficiency of steady-state target scanning no longerallows identification of each target cell with time periods of hours.However, even in this case viral outgrowth may potentially still belimited upon encounter of a small number of targets, through the“field effect” described by Müller et al. (22).Comparisons of the number of target cell encounters by these

simulated cells with that of simulated T cells that were eithernonmotile or nondendritic indicate that T-cell motility is theprimary factor that determines the efficiency of target cell en-counter, with dendricity adding another 33% (Fig. S4C). Asa side note, modeling also indicated that the dendritic (com-pared with amoeboid) morphology allowed migrating T cells totouch an 11% larger cell surface area of each contacted kerati-nocyte per time point (P < 0.001), a property that may con-ceivably increase the efficiency of scanning.

Skin Patrol by Tissue-Resident Memory T Cells Allows RapidRecognition of Antigen-Expressing Cells in Vivo. To directly testthe hypothesis that the epidermal migration by long-term residentmemory T cells leads to antigen scanning and thereby allows thedetection of antigen-expressing cells, we first established howantigen recognition by skin-resident memory T cells can be vi-sualized in vivo. To this purpose, sites containing a GFP+gBmemory T-cell pool were either left untreated or injected with thegB498–505 peptide, with control peptides (CMV pp65495–503 andOva257–264), or with vehicle. In all cases, skin-resident Tmemdisplayed a transient increase in circularity and a decrease inmigration speed following injection, likely reflecting skin trauma(Fig. 5 C and D and Fig. S5A). However, 3 h postinoculation,GFP+gB Tmem that had either been exposed to vehicle or tocontrol peptides had recovered their dendritic shape and migra-tory behavior (Fig. 5 C–E and Movie S7). In sharp contrast, gBTmem that had been exposed to gB498–505 peptide remainedroundish and arrested for up to 9 h after peptide delivery (Fig. 5C–E and Movie S8). This change in shape and motility was strictlyantigen-dependent, because skin-resident OTI Tmem underwentsimilar changes upon Ova257–264 delivery, but not in the presenceof pp65495–503 or gB498–505 control peptides (Fig. S5B). In an in-dependent set of experiments, s.c. peptide delivery was alsoshown to result in T-cell rounding up and stalling in an antigen-specific manner (Fig. S6). These data demonstrate that antigenrecognition by skin-resident Tmem can be visualized by alter-ations in both motility and morphology; the loss of motility isreminiscent of prior data on antigen recognition by naïve andeffector T cells (23), and the large change in morphology isa characteristic that is unique to skin-resident memory T cells.Having established both loss in dendricity and migration as

experimental readouts to determine whether migrating Tmemcan identify rare antigen-expressing cells, we created a situationin which only a small fraction of skin cells underwent de novoexpression of an antigen recognized by the skin-resident memoryT-cell population. To this purpose, we injected epidermis hostinga local GFP+gB memory T-cell pool with DNA encoding theKatushka fluorescent protein as a genetic fusion to the gB498–505peptide. Following DNA delivery, Katushka-gB498–505–expressingcells were readily detectable amid a large number of antigen-negative cells (Movie S9). Remarkably, GFP+gB Tmem could beshown to continue migration and dendrite protrusion through theskin until they contacted an antigen-expressing target cell, afterwhich they immediately stalled and rounded up (Fig. 6 A and B).Furthermore, stalling and rounding could also be observed formigrating Tmem even before antigen expression by the neighboringcells became detectable by microscopy (Fig. 6C and Movie S9),suggesting that recognition of antigen expression during naturalinfection may occur before substantial pathogen replication cantake place. Both stalling and rounding were fully antigen-de-pendent; Katushka-expressing skin cells that lacked the gB498–505epitope were completely ignored (Movie S9).

BA

C

- 1 0 3 6 35

GFP+OTI

vaccination

imaging

- 1 0 3 6 11

GFP+OTI

vaccination

imaging

HSV

34

HSV

10

Teff Tmem

0.00

0.25

0.50

0.75

1.00

1.25

circ

ular

ity

p< 0.0001

Tmem+ HSV

GFP+OTI Teff +HSV

GFP+OTI Tmem steady state skin

GFP+OTI Tmem + HSVD

Kaede+gB Tmem - nuclei of recipient’ skin cells

top/ bottom view bottom/ top view

Fig. 4. Aspects of local migration and dendricity of memory T cells. (A and B)Recipients of GFP+OTI T cells (day −1) were vaccinated with a DNA vaccineencoding the Ova257–264 peptide (days 0, 3, and 6). Subsequently, mice wereinfected with HSVTOM at day 10 postvaccination and bystander T cells weremonitored at the site of HSVTOM infection the next day (B, Upper); alternatively,micewere infectedwith HSVTOM at the site of formerDNA vaccination at day 34postvaccination and locally resident Tmem were monitored the next day (B,Lower). Images are confocal maximum intensity projections representative offive different experiments. (Scale bar, 50 μm.) (C) Circularity of epidermal ef-fector T cells and skin-resident Tmemwas determined frommaximum intensityprojections of recorded confocal Z-stacks, with a value of 1.0 reflecting a per-fectly circular shape. (D) Detail of a Tmem crawling in between epithelial cells.H-2B-GFP transgenic recipients of Kaede+gB T cells were vaccinated with a DNAvaccine encoding the gB498–505 epitope and infected with HSVTOM, according tothe scheme in Fig. 1A. One month postinfection, Kaede was photoconvertedand imaging was performed. Top and bottom view of the same flipped Z-stackare shown. Representative of two experiments. (Scale bar, 10 μm.)

19742 | www.pnas.org/cgi/doi/10.1073/pnas.1208927109 Ariotti et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

DiscussionTaken together, the data presented here demonstrate that theCD8+ memory T cells that remain behind in areas of prior in-flammation actively patrol the epidermis for antigen-expressingcells, independent of residual antigen or pathogen. The speedwith which these cells migrate is much lower than that of lymphnode T cells, perhaps explaining why this property has thus farnot been appreciated (2). Importantly, though, we here dem-onstrate that, because of their high directional persistence, skin-resident memory T cells do efficiently patrol large portions oftissue, within time scales that are relevant to the time scale ofviral replication. To carry out this patrol, tissue CD8+ Tmemextend dendrites in a 2D plane, thereby contrasting their

morphology to that of sessile epidermal γδ T cells and Langer-hans cells. Local memory T-cell patrol is likely to be of particularvalue in cases in which a tissue is prone to multiple attacks fromthe same infectious agent, such as during reactivation of herpessimplex or varicella zoster viruses (24). Thus, skin-resident CD8+

memory T cells can be considered an active first line of defensefor secondary infections: Whereas during primary immuneresponses CD8+ T cells await the delivery of antigens that arebrought to the draining lymph nodes by APCs, during localsecondary responses in skin, memory T cells proactively searchfor antigen themselves. This observation raises the intriguingpossibility that skin-resident memory T cells may be independentof local APCs for their protective function.

steady state

OVA(257-264)

gB(498-505)

pp65(495–503)

B

C shape

t= 0’

D

hours post peptide injection

0 10 20 300

500

1000

1500

gBT; gB

gBT; pp65

gBT; Ova

steady state

OT-I; Ova

track duration (min)

mea

n sq

uare

disp

lace

men

t (µm

2 )

t= 0’ 30’

0 3 6 9 0 3 6 9 0 3 6 9 0 3 6 90.2

0.4

0.6

0.8

1.0

circu

larit

y

37 Tmem

93 Tmem187 Tmem430 Tmem

0 4 8 120

20

40

60

80

100

time to find target (hours)

% o

f fou

nd ta

rget

A

gBT; pp65gBT; Ova

gBT; gBOT-I; Ova

dendritic motile

motility

E

roundish motile

dendritic still

Fig. 5. Dendricity and motility of skin-resident memory T cells allow skin patrol and are regulated by TCR triggering. (A) Simulations (CPM) of Tmem in a fieldof keratinocytes were performed for 4 h in “memory T-cell time” (i.e., with the in silico cells covering a similar distance as Tmem in experimental meas-urements). Examples of images at the final time point of simulation for dendritic-motile simulated T cells (with motility, persistence, and dendricity matchedto the experimental Tmem), roundish-motile simulated T cells (with motility and persistence matched to the experimental Tmem, but with the circular ap-pearance of effector T cells), and dendritic-still simulated T cells (nonmotile cells that extend and retract dendrites) in a field of 500 × 500 pixels. Keratinocytescontacted by simulated T cells are depicted in different shades of blue, with darker blue indicating more intense contact (based on cumulative contacted areaof the skin cell). (B) The time to find antigen-expressing “targets” in silico, using either the experimentally observed densities of simulated T cells (230 T cells/field, equivalent to 430 T cells/mm2) or the indicated lower densities. (C) Time-lapse imaging of GFP+gB T cells at 9 h after peptide injection. Recipients of GFP+

gB cells were infected with HSVTOM. Three months after local infection, the skin was injected with gB498–505 peptide, CMV pp65495–503 peptide, Ova257–264peptide, or left untreated. Time-lapse imaging of the area was conducted for 30 min at 0, 3, 6, and 9 h after peptide delivery. Left: initial image of each time-lapse. (Right) Superimposition of all images comprising the 30-min time-lapse imaging session. Representative of three experiments. (Scale bar, 50 μm.) (D)Mean ± SEM circularity of the indicated Tmem was determined by imaging at the indicated time points after peptide delivery. Dashed line indicates meancircularity of skin-resident memory T cells in untreated animals (see also Fig. 3C). (E) Tmem motility, depicted as a mean square displacement ± SEM plot, asdetermined by imaging at 9 h after peptide delivery.

Ariotti et al. PNAS | November 27, 2012 | vol. 109 | no. 48 | 19743

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

At present, it remains unclear which factors determine whetherskin-resident T-cell memory is only formed locally or throughoutthe skin surface (2, 9). Nevertheless, the observation that locallypatrolling T-cell memory can be induced by vaccination (Fig. 1 A

and B) but also at inflamed sites in which cognate antigen isabsent (6; Fig. 2 A and B) indicates that it should be possible tosteer the formation of T-cell memory to defined epithelial areas.Based on the known protection provided by local immunity, thiscould be used to provide local immune defense against virusesthat enter the body at well-defined epithelial sites.

Materials and MethodsAdoptive Transfer, HSV Infection, Peptide, and DNA Injection. Spleens fromGFP+gBT I.1 or GFP+OTI mice were isolated and homogenized; after redblood cell lysis, CD8+ T cells were isolated and C57BL/6j mice were injectedi.v. with 2.0 × 105 sorted naïve CD8+ T cells. Intraepithelial HSV infection wascarried out on the shaved flank of anesthetized mice by tattooing a dropletcontaining ∼2.5 × 105 pfu HSVTOM onto the skin using a sterile disposable 11-needle bar mounted on a rotary tattoo device. For peptide delivery, 750 ngof peptide in PBS/10% (vol/vol) DMSO solution was tattooed in the flank ofanesthetized mice. For skin cell transfection, 30 μg of the indicated DNA inPBS was applied onto the skin and injected using a sterile disposable 11-needle bar mounted on a rotary tattoo device (25).

Confocal Microscopy. Anesthetized mice were placed in a custom-builtchamber with the infected flank gently placed against a coverslip at thebottom side of the chamber. Images were acquired using an inverted LeicaTCS SP2 confocal scanning microscope equipped with diode and argon lasersand enclosed in a custom-built environmental chamber that was maintainedat 37 °C using heated air. Images were acquired using a 20×/0.7 N.A. dryobjective. GFP was excited at 488 nm wavelength and td-tomato at 561 nm.Typical voxel dimensions were 0.7–1.5 μm laterally × 1.5–2.5 μm axially.Three-dimensional stacks (typical size ∼500 μm × ∼500 μm × ∼30 μm) werecaptured every 2 min for a period of up to 4 h.

ACKNOWLEDGMENTS.We thank the NKI Flow Cytometry Facility and DigitalMicroscopy Facility for technical support and Ruud van Mierlo for assistance.This work was supported by The Netherlands Organization for ScientificResearch Grant 912.10.066 (to T.N.S. and R.J.d.B.), Grants 917.10.330 and175.010.2007.00 (to J.v.R.), Veni Grant 916.86.080 (to J.B.B.), and EuropeanResearch Council Grant Life-his-T (to T.N.S.).

1. Bevan MJ (2011) Memory T cells as an occupying force. Eur J Immunol 41(5):1192–1195.

2. Gebhardt T, et al. (2011) Different patterns of peripheral migration by memory CD4+and CD8+ T cells. Nature 477(7363):216–219.

3. 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.

4. Wakim LM, Woodward-Davis A, Bevan MJ (2010) Memory T cells persisting within thebrain after local infection show functional adaptations to their tissue of residence.Proc Natl Acad Sci USA 107(42):17872–17879.

5. Masopust D, et al. (2010) Dynamic T cell migration program provides resident memorywithin intestinal epithelium. J Exp Med 207(3):553–564.

6. Mackay LK, et al. (2012) Long-lived epithelial immunity by tissue-resident memoryT (TRM) cells in the absence of persisting local antigen presentation. Proc Natl AcadSci USA 109(18):7037–7042.

7. Ariotti S, Haanen JB, Schumacher TN (2012) Behavior and function of tissue-residentmemory T cells. Adv Immunol 114:203–216.

8. Mintern JD, Guillonneau C, Carbone FR, Doherty PC, Turner SJ (2007) Cutting edge:Tissue-resident memory CTL down-regulate cytolytic molecule expression followingvirus clearance. J Immunol 179(11):7220–7224.

9. Jiang X, et al. (2012) Skin infection generates non-migratory memory CD8+ T(RM)cells providing global skin immunity. Nature 483(7388):227–231.

10. Cose SC, Kelly JM, Carbone FR (1995) Characterization of diverse primary herpessimplex virus type 1 gB-specific cytotoxic T-cell response showing a preferential V betabias. J Virol 69(9):5849–5852.

11. Miller MJ, Wei SH, Parker I, Cahalan MD (2002) Two-photon imaging of lymphocytemotility and antigen response in intact lymph node. Science 296(5574):1869–1873.

12. Nishibu A, et al. (2006) Behavioral responses of epidermal Langerhans cells in situ tolocal pathological stimuli. J Invest Dermatol 126(4):787–796.

13. Kubo A, Nagao K, Yokouchi M, Sasaki H, Amagai M (2009) External antigen uptake byLangerhans cells with reorganization of epidermal tight junction barriers. J Exp Med206(13):2937–2946.

14. Chodaczek G, Papanna V, Zal MA, Zal T (2012) Body-barrier surveillance by epidermalγδ TCRs. Nat Immunol 13(3):272–282.

15. Harris TH, et al. (2012) Generalized Lévy walks and the role of chemokines inmigration of effector CD8+ T cells. Nature 486(7404):545–548.

16. Tomura M, et al. (2008) Monitoring cellular movement in vivo with photoconvertiblefluorescence protein “Kaede” transgenic mice. Proc Natl Acad Sci USA 105(31):10871–10876.

17. Beltman JB, Marée AF, de Boer RJ (2007) Spatial modelling of brief and long inter-actions between T cells and dendritic cells. Immunol Cell Biol 85(4):306–314.

18. Glazier JA, Graner F (1993) Simulation of the differential adhesion driven re-arrangement of biological cells. Phys Rev E Stat Phys Plasmas Fluids Relat InterdiscipTopics 47(3):2128–2154.

19. Graner F, Glazier JA (1992) Simulation of biological cell sorting using a two-dimensional extended Potts model. Phys Rev Lett 69(13):2013–2016.

20. Beltman JB, Marée AF, Lynch JN, Miller MJ, de Boer RJ (2007) Lymph node topologydictates T cell migration behavior. J Exp Med 204(4):771–780.

21. Mulholland WJ, et al. (2006) Multiphoton high-resolution 3D imaging of Langerhanscells and keratinocytes in the mouse skin model adopted for epidermal powderedimmunization. J Invest Dermatol 126(7):1541–1548.

22. Müller AJ, et al. (2012) CD4+ T cells rely on a cytokine gradient to control intracellularpathogens beyond sites of antigen presentation. Immunity 37(1):147–157.

23. Mempel TR, Henrickson SE, Von Andrian UH (2004) T-cell priming by dendritic cells inlymph nodes occurs in three distinct phases. Nature 427(6970):154–159.

24. Diefenbach RJ, Miranda-Saksena M, Douglas MW, Cunningham AL (2008) Transportand egress of herpes simplex virus in neurons. Rev Med Virol 18(1):35–51.

25. Bins AD, et al. (2005) A rapid and potent DNA vaccination strategy defined by in vivomonitoring of antigen expression. Nat Med 11(8):899–904.

contact no contact

0.00

0.25

0.50

0.75

1.00p = 0.029

(n = 8) (n = 34)

circ

ular

ity

contact

no contact

A B

*

**

<< <

C 1h26’ 3h16’2h20’

Fig. 6. Patrolling skin-resident memory T cells can quickly identify raretargets. (A) Cell migration tracks of individual GFP+gB Tmem either in con-tact or not in contact with an antigen-expressing target, normalized for theirorigin. (B) Circularity of GFP+gB Tmem was analyzed from maximum in-tensity projections of individual cells. Cells were categorized based on con-tact with Katushka-gB498–505–expressing cells at any time point during theimaging session, derived from one experiment. Bars indicate mean ± SEM.(C) Sequential images of migrating skin-resident memory T cells that eitherstop upon contact with specific target (asterisk) or keep migrating withouttarget contact (arrow).

19744 | www.pnas.org/cgi/doi/10.1073/pnas.1208927109 Ariotti et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0