LETTERS

Tyrosine kinase receptor RET is a key regulator ofPeyer’s Patch organogenesisHenrique Veiga-Fernandes1, Mark C. Coles1, Katie E. Foster1, Amisha Patel1, Adam Williams1, Dipa Natarajan2,Amanda Barlow2, Vassilis Pachnis2 & Dimitris Kioussis1

Normal organogenesis requires co-ordinate development andinteraction of multiple cell types, and is seemingly governed bytissue specific factors. Lymphoid organogenesis during embryoniclife is dependent on molecules the temporal expression of which istightly regulated. During this process, haematopoietic ‘inducer’cells interact with stromal ‘organizer’ cells, giving rise to thelymphoid organ primordia1. Here we show that the haemato-poietic cells in the gut exhibit a random pattern of motility beforeaggregation into the primordia of Peyer’s patches, a major com-ponent of the gut-associated lymphoid tissue. We further showthat a CD451CD42CD32Il7Ra2c-Kit1CD11c1 haematopoieticpopulation expressing lymphotoxin has an important role in theformation of Peyer’s patches. A subset of these cells expresses thereceptor tyrosine kinase RET, which is essential for mammalianenteric nervous system formation2. We demonstrate that RET sig-nalling is also crucial for Peyer’s patch formation. Functional gen-etic analysis revealed that Gfra3-deficiency results in impairmentof Peyer’s patch development, suggesting that the signalling axisRET/GFRa3/ARTN is involved in this process. To support thishypothesis, we show that the RET ligand ARTN is a strong attract-ant of gut haematopoietic cells, inducing the formation of ectopicPeyer’s patch-like structures. Our work strongly suggests that theRET signalling pathway, by regulating the development of boththe nervous and lymphoid system in the gut, has a key role in themolecular mechanisms that orchestrate intestine organogenesis.

The formation of Peyer’s patches is dependent on the colonization ofthe gut wall during embryogenesis by haematopoietic lymphoid tissueinducer (LTi) cells which, following interactions with resident stromalymphoid tissue organizer (LTo) cells, aggregate on embryonic day E16.5to generate the Peyer’s patch primordia1,3–5. Deficiency in Ikaros6, RORc(ref. 7), Id2 (ref. 8), TRANCE/TRANCER9–11 and CXCL13–CXCR5 (refs12, 13) result in developmental or functional deficits of LTi cells and animpairment in Peyer’s patch formation, whereas lymphotoxin b recep-tor (LTbr)-deficient mice fail to form Peyer’s patches owing to impairedLTo stroma cell function14–17. Il7–Il7Ra interactions have also beenimplicated in Peyer’s patch formation4,5,18,19.

Human CD2–GFP transgenic mice express green fluorescent pro-tein (GFP) in a population of haematopoietic cells aggregating inembryonic sites that develop lymph nodes in the adult20. Such cellswere also detected in the gut by fluorescence microscopy from dayE12.5 (Fig. 1a). Fluorescent cells were initially restricted to the pyl-oric/duodenum region and the caecum, but were found evenly dis-tributed throughout the gut by stage E15.5 (Fig. 1b). Within the next24 h these cells aggregate to form Peyer’s patch primordia (Fig. 1b).To examine whether the final number of haematopoietic cells withinthe gut is achieved by the expansion of an early precursor pool or thecontinuous influx of new cells, we analysed the cell cycle profile ofthe enteric GFP1 cells. Virtually all CD451GFP1 cells from E15.5 or

1Division of Molecular Immunology, 2Division of Molecular Neurobiology, MRC National Institute for Medical Research, The Ridgeway Mill Hill, London NW7 1AA, UK.

E12.5 E13.5 E14.5

Caecum

Mid-gut

a

E15.5 E16.5b

Mid-gut

StomachS

SS

C C C

P/D

P/D P/D

c

G0/G1 S G2+M

Cel

lco

un

tsDNA

E16.5

0 min 8 min 16 min

d

0 50 1000

50

100

X (µm)

Y (µ

m)

e

3.7 µm min–1

Figure 1 | Colonization of the embryonic intestine by GFP1 haematopoieticcells. a, Human CD2–GFP1 embryos were dissected and the intestinesanalysed by stereo fluorescence microscopy (31.6 objective). Results show thedistribution of GFP1 cells from E12.5 to E14.5 in the stomach (S), pylorus/duodenum junction (P/D), caecum (C) and mid-gut. b, GFP1 cells in the mid-gut from E15.5 to E16.5 are depicted showing Peyer’s patch primordiumformation by E16.5. c, E16.5 GFP1 Peyer’s patch primordia were micro-dissected and digested with collagenase. Cell suspensions were sorted by flowcytometry for GFP1 cells and their cell cycle status analysed by propidiumiodide incorporation. GFP1 cell counts (2,063) are shown in the histogram.Similar results were obtained at E15.5. d, E15.5 GFP1 intestines wereimmobilized in a collagen matrix and time-lapse stereo microscopy analysiswas performed. Results show detail of time-lapse analysis showing differenttime points and relative positions of two different cells during the analysis.e, Cell tracks are shown for six different cells analysed over 25 min in the sameconditions as in d. Average cell speed is indicated (3.7 6 0.6mm min21; error iss.d.). For the time-lapse sequence please see Supplementary Video 1 and 2.

Vol 446 | 29 March 2007 | doi:10.1038/nature05597

547Nature ©2007 Publishing Group

E16.5 were arrested at the G0/G1 stage of the cell cycle (Fig. 1c),consistent with the idea that proliferation does not have a major rolein the increase of the enteric CD451GFP1 pool during embryogen-esis. However, these cells exhibit a remarkable and seemingly randommotility, often reaching 6mm min21 (average speed 3.7 mm min21;Fig. 1d, e; Supplementary Video 1 and 2), suggesting that new influxand migration are the main factors for their dissemination within thegut wall.

Phenotypic analysis of cells from days E15.5 and E16.5 revealedthat the CD451GFP1 population included the CD41CD32Il7Ra1

c-Kit1 LTi cells described previously1. In addition, we detecteda phenotypically distinct population of CD42CD32Il7Ra2c-Kit1

CD11c1 cells, which followed a pattern of distribution and aggrega-tion like that of the LTi cells and are probably similar to the CD11c1

cells previously described in E17.5 intestine3,21 (Fig. 2a, b; Sup-plementary Fig. 1a–d; Supplementary Video 3).

Phenotypic analysis revealed differences and similarities betweenthe CD42CD32Il7Ra2c-Kit1CD11c1 cells from the embryonic gutand dendritic cells isolated from adult spleens. Thus, CD42CD32

Il7Ra2c-Kit1CD11c1 cells express major histocompatibility com-plex class II and CD11b similar to the splenic dendritic cells, but lackthe DEC205 marker and are positive for the surface markers GR-1and NK1.1, which are not expressed by the adult dendritic cells(Supplementary Fig. 1b). Furthermore, we found that comparedwith CD41CD32Il7Ra1c-Kit1 LTi cells, CD42CD32Il7Ra2c-Kit1

CD11c1 cells express higher levels of lymphotoxin b (Fig. 2b; Sup-plementary Fig. 3d), a molecule known to be crucial for Peyer’s patchorganogenesis1,14–17,22,23.

To determine whether these cells have a role in the formation ofPeyer’s patches, we examined transgenic mice that express the diph-theria toxin receptor (DTR) and GFP under the control of the CD11cpromoter (CD11c-GFP–DTR mice)24. These mice express DTR andGFP in approximately 20–30% of the CD11c cells in E14.5 to E17.5gut (Supplementary Fig. 2a). Administration of diphtheria toxin toCD11c-GFP–DTR transgenic embryos resulted in a significant re-duction (25%) in the number of Peyer’s patches (Fig. 2c). Takentogether, these experiments suggest that the formation of normalnumbers of Peyer’s patches during gut organogenesis requires thefull complement of CD42CD32Il7Ra2c-Kit1CD11c1 cells.

The colonization of the embryonic gut by lymphoid cells and theiraggregation to form Peyer’s patches is reminiscent of the invasion ofthe gut mesenchyme by neuroectodermal cells that colonize the gut

a

E15.5 E16.5

GFP+

CD

4

CD11c

37.8

16.1

39

21.3

E18.5

0

5

10

PBS DT

P = 0.0037

PBS DT

WT

c

P = 0.28

b

c-Kit

LTβ

91.7 59.8

20 43

GFP+CD11c+GFP+CD4+

IL7Rα

94.5 0.2

E15.5

Num

ber

of P

eyer

’s

pat

ches

CDIIc-DTR–GFP

Figure 2 | Phenotype of embryonic GFP1 haematopoietic cells in theintestine. a, GFP1 haematopoietic cells from E15.5 and E16.5 intestineswere purified after collagenase treatment. Cell suspensions were analysed byflow cytometry analysis and dot plots depicted. Numbers represent cellpercentages in the respective gates. b, E15.5 intestines were analysed asabove. Left panels, expression profiles on GFP1CD41 cells; right panels,expression profiles on GFP1CD11c1 cells. Dotted lines show negativecontrol staining. GFP1 cells are also CD32CD192CD8a2 (SupplementaryFig. 1a). c, CD11c-DTR–GFP transgenic mice were crossed with humanCD2–DsRed transgenic mice (Supplementary Fig. 2c) and were treated daily(intraperitoneal injection) from E14.5 to E17.5 with 4 nanograms ofdiphtheria toxin (DT; Sigma) per gram of mouse. At E18.5 Peyer’s patcheswere counted by stereo fluorescent microscopy (Supplementary Fig. 2c).Results show Peyer’s patch numbers in CD11c-DTR–GFP and wild-type(WT) mice treated with phosphate-buffered saline (PBS) or DT. CD11c-DTR–GFP PBS-treated, n 5 6; CD11c-DTR–GFP DT-treated, n 5 22; wildtype PBS-treated, n 5 3; wild type DT-treated, n 5 14; two-tailed t-test Pvalues were CD11c-DTR–GFP PBS-treated versus CD11c-DTR–GFP DT-treated, P 5 0.0037; wild-type PBS-treated versus wild type DT-treated,P 5 0.279; CD11c-DTR–GFP DT-treated versus wild type DT-treated,P 5 0.043. Error bars show s.d.

b

c

3.5

3.4CD

45

CD45+

FSC Il7RαCD11c

WT

Ret–/–

E17.5

WT Ret–/–

eE17.5

WT Ret 51/51

TC

Rαβ

CD8α

0

4

8

12

Adult

WT Ret 51/51

d

P = 0.004

16.5

16.9

4

4

WT

Ret 51/51

CD

4

12

18

13

17

12.1

13.1

GFP+CD11c+

Cel

lco

un

ts

GFP+CD4+

RET

40.30.7

a

Num

ber

of P

eyer

’s

pat

ches

Figure 3 | Peyer’s patch development in tyrosine kinase RET mutants.a, Expression of tyrosine kinase RET in the intestine by E14.5. Results showRET expression on GFP1CD41 (left) and GFP1CD11c1 (right) cells.Dotted lines show negative control staining. No other RET1 cells weredetected among GFP1 cells (Supplementary Fig. 3b); CD11c1RET1 positivecells do not express Sox10 or Mash1, which are highly expressed in neuralcrest cells (not shown). b, E16.5 and E17.5 intestines from RET-deficientembryos (Ret2/2) and wild-type (WT) littermate controls were fixed andwhole-mount stained with monoclonal CD4 antibodies. Whole-mountstainings were analysed by stereo and confocal microscopy (310 objective)for Peyer’s patch development. Confocal images of mid-gut from wild-typeand Ret2/2 embryos are shown. Grey, intestine structure; green, CD4. Thesame patterns were obtained in six independent experiments. Number ofanalysed embryos were: wild type and Ret1/2, n 5 18; Ret2/2, n 5 10. Errorbars show s.d. c, E15.5 intestines from Ret2/2 and wild-type littermatecontrols were purified as in Fig. 2a and cell suspensions analysed by flowcytometry analysis. CD4, CD11c and Il7Ra expression is shown on CD451

haematopoietic gated cells. d, Intestines from Ret51/51 adults and wild-typelittermate controls were analysed by stereo microscopy and Peyer’s patchescounted. Left panels, number of Peyer’s patches in Ret51/51 and wild-typemice. Wild type, n 5 5; Ret51/51, n 5 5; two-tailed t-test P 5 0.004. Error barsshow s.d. Right panels, flow cytometry analysis of Peyer’s patches fromRet51/51 and wild-type adult mice. e, E17.5 intestines from Ret51/51 embryosand wild-type littermate controls were analysed as in Fig. 3b. Grey, intestinestructure; green, CD4.

LETTERS NATURE | Vol 446 | 29 March 2007

548Nature ©2007 Publishing Group

and coalesce to form enteric ganglia. Because the receptor tyrosinekinase RET has a key role in the formation of the mammalian entericnervous system2, we examined whether CD451GFP1 cells in the gutlymphoid tissue primordia also express this receptor. Using a mono-clonal antibody against RET we show that ,40% of the CD42CD32

Il7Ra2c-Kit1CD11c1 cells found in the embryonic gut are also pos-itive for this receptor (Fig. 3a). These results were further confirmedby indirect immunofluorescence of whole-mount gut preparationsand RNA expression (Supplementary Figs 1d and 3a).

To assess the functional role of RET expressed in CD42CD32

Il7Ra2c-Kit1CD11c1cells of the gut, we examined formation ofPeyer’s patches primordia in the intestines of E16.5 and E17.5embryos homozygous for a severe loss-of-function mutation in Ret(Ret2/2)2. Compared with wild-type or heterozygous (Ret1/2) lit-termates, E17.5 Ret2/2 embryos show an absence of Peyer’s patchprimordia (Fig. 3b). Despite the absence of Peyer’s patches, CD41

LTi cells were detectable in the gut wall of Ret2/2 embryos at thisstage (Fig. 3b). Moreover, fluorescence-activated cell sorting (FACS)analysis of haematopoietic cells from the gut of E15.5 Ret2/2 andwild-type littermates revealed no differences in the relative frequencyor phenotype of the recovered cells (Fig. 3c), indicating that Ret2/2

mice have a full complement of gut LTi cells. We therefore suggestthat the primary defect in Peyer’s patch primordia formation inRet2/2 mice is not due to the absence of haematopoietic subpopula-tions, but rather to their failure to aggregate.

To examine whether the absence of Peyer’s patches in Ret2/2

mutants is secondary to the failure of enteric nervous system forma-tion in these embryos, we analysed mice homozygous for a hypo-morphic allele of Ret (Ret51), which results in distal colonic agang-lionosis, but an apparently normoganglionic small intestine (wherePeyer’s patches are located)25. Despite the presence of a normalcomplement of enteric ganglia in the small intestine, the number ofPeyer’s patches in adult Ret51/51 animals is significantly reduced com-pared with wild-type or heterozygous (Ret1/51) littermates (Fig. 3d),suggesting that the Peyer’s patches phenotype observed in Ret2/2

animals is not secondary to enteric nervous system deficits. Peyer’spatches in Ret51/51 animals formed at approximately the right timeduring embryogenesis (Fig. 3e) and had normal size, their overallorganization and cell type content was similar to that in controlanimals (Fig. 3d; Supplementary Fig. 4). These results argue againstthe idea that the absence of Peyer’s patches in Ret2/2 mice is due to adelay in the timing of Peyer’s patch primordia formation and suggestthat RET signalling has a crucial role in the early events of nucleationof the Peyer’s patch primordium rather than controlling later eventsrelated to the organization and maintenance of the structure.

Previous studies have established that neural crest proliferation andmigration within the gut wall depends on the cell autonomous activa-tion of RET by stroma-derived glial-cell-line-derived neurotrophicfactor (GDNF) and its co-receptor GFRa1 (refs 26–28). Perinatal micehomozygous for null alleles of Gdnf (Gdnf2/2) and Gfra1 (Gfra12/2)showed that neither GDNF nor GFRa1 are implicated in Peyer’s patchformation, because mice deficient in these molecules develop a normalcomplement of Peyer’s patches (Supplementary Fig. 5). In addition,these studies provide further evidence that RET signalling is indepen-dently required within the neural crest and lymphoid lineages. Ex-pression analysis of the genes encoding the other RET ligands revealedthat both Artn and Nrtn (but not Pspn) were expressed by non-haematopoietic stroma cells within the gut wall (Fig. 4a; Supplemen-tary Fig. 3a). Importantly, Artn was expressed in VCAM11 gut stromalcells, which include the LTo and blood vessel endothelial cells (Fig. 4a).Because it has been previously shown that ARTN is not produced byendothelial cells29, this finding indicates that the VCAM11 LTo cellsmay be the main source of the ligand that is responsible for RETsignalling in gut lymphoid tissue. Assessing Peyer’s patch formationin mice homozygous for a null mutation in Gfra3 (Gfra32/2), the locusencoding the main co-receptor for ARTN, showed that Gfra32/2 ani-mals have significantly fewer, albeit well-formed, Peyer’s patches

(Fig. 4b, c; Supplementary Fig. 4), suggesting that the ARTN/GFRa3/RET signalling axis is involved in gut lymphoid organogenesis.However, because the phenotype observed in Gfra32/2 mice is partial,it is likely that in the absence of GFRa3, other ligands (NRTN) andco-receptors (GFRa2) that are expressed in the microenvironment ofCD42CD32Il7Ra2c-Kit1CD11c1RET1 cells, may also contribute toPeyer’s patch primordia formation.

Further evidence that ARTN/RET interactions have a role inthe mobility and aggregation of GFP haematopoietic cells was pro-vided by in vitro organ culture experiments. When whole embryonicgut was incubated with agarose beads soaked in ARTN, large aggre-gates of GFP1 cells were formed in the vicinity of the beads (Fig.4d, e). Remarkably, immunohistochemical staining of these aggre-gates revealed that, in addition to the expected CD11c1 cells, a largenumber of CD41 LTis were also found within these aggregates(Fig. 4e). Furthermore, upregulation of VCAM1 was also observedin presumably sessile mesenchymal cells in the vicinity of the cellularaggregates (Fig. 4e), altogether giving rise to an ectopic Peyer’s patchprimordium structure. Importantly, these ectopic structures werelikely to be the exclusive product of GFP1 cell migration, because

Artn

Nrtn

CD4+

CD11c+

VCAM1+

VCAM1–GFP+ GFP–a

Hprt

cE17.5

WT Gfra3

dBSA ARTN

24 h

48 h

72 h

Adult

TC

Rαβ

CD8α

8.8

9.8

2

3

WT

Gfra3–/–

–/–

WT Gfra3–/–

b

0

4

8

12

Gfra3

CD4CD11c

GFPVCAM1

eGFP+ cell aggregates

B

P = 0.003

Num

ber

of P

eyer

’s

pat

ches

Figure 4 | RET/GFRa3/ARTN signalling axis and Peyer’s patchdevelopment. a, Cells from E15.5 intestines were purified as in Fig. 2a. Cellsuspensions were sorted by flow cytometry for different cell populations andexpression analysis carried out by RT–PCR. b, Intestines from Gfra32/2

adults and wild-type littermate controls were analysed by stereo microscopyand Peyer’s patches counted. Left panel, number of Peyer’s patches inGfra32/2 and wild-type mice. Wild type, n 5 10; Gfra32/2, n 5 8; two-tailedt-test, P 5 0.003. Error bars show s.d. Right panels, flow cytometry analysisof Peyer’s patch cells from Gfra32/2 and wild-type adult mice. c, E17.5intestines from Gfra32/2 embryos and wild-type littermate controls wereanalysed as in Fig. 3b. Grey, intestine structure; green, CD4. d, E15.5 GFP1

intestines were cultured in vitro with agarose beads impregnated with BSA(left) or ARTN recombinant protein (right). Analysis was performed over72 h by stereo microscopy (31.6 objective). Arrows show the sameindividual bead over time. Similar results were obtained in more than fiveindependent experiments. e, GFP1 cell aggregates were obtained asdescribed in Fig. 4d. Images were obtained using confocal microscopy (340objective). Left panel, CD4 (green) and CD11c (red) staining. B, agarosebead. Right panel, GFP (green) and VCAM1 (red).

NATURE | Vol 446 | 29 March 2007 LETTERS

549Nature ©2007 Publishing Group

no GFP1BrdU1 cells were detected in this assay (SupplementaryFig. 6).

Our data demonstrate that RET signalling is required duringembryogenesis for the development of secondary lymphoid struc-tures of the gut and provide evidence that ARTN/GFRa3 is at leastone of the critical regulators of RET-dependent Peyer’s patch prim-ordia formation. Our data also suggest that RET is required for theproper migratory behaviour of haematopoietic cells during embryo-genesis. This suggestion is based on the observed motility of entericGFP1 cells and the ability of ARTN to promote the formation of largeectopic aggregates of GFP1 expressing cells in gut explants.

It is possible that RET activation in CD42CD32Il7Ra2

c-Kit1CD11c1RET1 cells may result in direct or indirect lympho-toxin b production (Fig. 2b) leading to the stimulation of LTo cellsthrough their LTbR. Likewise, the ability to produce and respond tochemokines might also be modulated by RET signalling. The resultspresented here are consistent with the hypothesis that signalling viaRET activates a gene expression programme empowering the haema-topoietic cells to respond to and affect the rest of the cells requiredfor Peyer’s patch development, including other LTi and LTo cells.Interestingly, B and T lymphocytes have been also reported to expressRET30, but the functional significance of this expression is unclear.

We propose that RET/ligand interactions are part of the initiatingevent that leads to the subsequent formation of the Peyer’s patchprimordium. In this context, the aggregation is the result of a cascadeof events triggered by the chance interaction of the randomly movingCD42CD32Il7Ra2c-Kit1CD11c1RET1 cells with stroma elementsleading to the attraction of CD41CD32Il7Ra1c-Kit1 LTi cells andsubsequent organization into secondary lymphoid organ primordia.The frequency of such an event occurring would be dependent on thenumbers of the different cellular participants and their motility, aswell as on the concentration of the different soluble factors. Such amodel is consistent with the fact that Peyer’s patches do not form infixed sites along the gut, whereas their number is fairly constant innormal mice. Because the aggregation of the cells is taking placewithin 24 h (between E15.5 and E16.5), it is very likely that the onsetof expression of a particular receptor or ligand, or the appearance (ormaturation) of a cellular component during that period is the rate-limiting step. Our results indicate that CD42CD32Il7Ra2

c-Kit1CD11c1 cells may have a decisive role in this rate-limitingstep, because their partial ablation results in a reduction of Peyer’spatch formation and deficiency in RET (expressed by these cells)results in absence of Peyer’s patches.

The findings presented in this paper provide evidence that molecu-lar mechanisms that have been historically ascribed to one specifictissue development may be used more generally to orchestrate theglobal development of organs containing structures derived fromdiverse germ layers. In the present case, the RET signalling pathwaythought to be essential for the development of the enteric nervoussystem is now revealed also to be critical for the formation of anenteric haematopoietic organ such as Peyer’s patches. Thus, by regu-lating both developmental processes RET emerges as a key moleculein the orchestration of intestine organogenesis.

METHODSMice. Human CD2–GFP transgenic mice31, CD11c-DTR–GFP transgenic mice24,

Ret2/2 B10 (ref. 2), Ret51/51 (ref. 2), Gfra12/2 (ref. 28), Gdnf2/2 (ref. 27) and

Gfra32/2 (ref. 32) mice were previously described. All mice were bred at NIMR

animal facilities.

Stereo and time-lapse microscopy. Stereo microscopy was performed using a

Zeiss M2Bio (Carl Zeiss) stereo-fluorescent microscope. Pictures were taken

using an Orca ER (Hammamatsu) camera and Openlab software (Improvi-

sion). Peyer’s patches of adult mice were counted using a 1.63 objective because

the size of these structures, around 2 mm, allows a clear identification at

such low magnifications. Analysis of GFP expression was done using a green

filter cube (emission, 525 nm) (Krammer scientific). Time-lapse microscopy

was performed using the above specified stereo fluorescence microscope. Still

pictures were taken every 30 s for 25 min. Movie sequences and cell tracking was

done using Volocity software (Improvision).

Confocal microscopy and whole-mount staining. Embryonic intestines were

micro dissected at different developmental stages using a stereo microscope.

Intestines were fixed in 4% PFA at room temperature for 20 min and then stained

using antibodies listed in Supplementary Methods 1. Samples were optically

cleared in BABB (Sigma) and acquired on a Leica SP2 microscope (Leica micro-

systems) using a 103/0.4 numerical aperture objective lens. Serial optical sec-

tions from whole-mount stained small intestines were taken. Three-dimensional

reconstruction of images was achieved using Volocity software (Improvision)

and snapshot pictures were obtained from the three-dimensional images.

Cellular analysis, flow cytometry and RT–PCR analysis. Embryonic intestines

were micro-dissected at different developmental stages using a stereo micro-

scope and further digested with collagenase H (Roche). Reverse transcription

and PCR (RT–PCR) reactions were carried out using reagents described else-

where33. Detailed information on protocols and reagents can be found in

Supplementary Methods 1 and 2.

In vitro aggregation assays. E15.5 intestines were micro-dissected from human

CD2–GFP mice. Whole intestines were cultured in vitro at 37 uC and incubated

with agarose beads impregnated overnight with 200 ng ml21 of recombinant

ARTN (PrepoTech EC) or BSA (Sigma). Pictures were taken using the stereo

microscope described above.

Received 26 October 2006; accepted 15 January 2007.Published online 25 February 2007.

1. Mebius, R. E. Organogenesis of lymphoid tissues. Nature Rev. Immunol. 3,292–303 (2003).

2. Schuchardt, A., D’Agati, V., Larsson-Blomberg, L., Costantini, F. & Pachnis, V.Defects in the kidney and enteric nervous system of mice lacking the tyrosinekinase receptor Ret. Nature 367, 380–383 (1994).

3. Adachi, S., Yoshida, H., Kataoka, H. & Nishikawa, S. Three distinctive steps inPeyer’s patch formation of murine embryo. Int. Immunol. 9, 507–514 (1997).

4. Adachi, S. et al. Essential role of IL-7 receptor alpha in the formation of Peyer’spatch anlage. Int. Immunol. 10, 1–6 (1998).

5. Yoshida, H. et al. IL-7 receptor a1 CD32 cells in the embryonic intestine inducesthe organizing center of Peyer’s patches. Int. Immunol. 11, 643–655 (1999).

6. Wang, J. H. et al. Selective defects in the development of the fetal and adultlymphoid system in mice with an Ikaros null mutation. Immunity 5, 537–549(1996).

7. Kurebayashi, S. et al. Retinoid-related orphan receptor c (RORc) is essential forlymphoid organogenesis and controls apoptosis during thymopoiesis. Proc. NatlAcad. Sci. USA 97, 10132–10137 (2000).

8. Yokota, Y. et al. Development of peripheral lymphoid organs and natural killercells depends on the helix–loop–helix inhibitor Id2. Nature 397, 702–706 (1999).

9. Kim, D. et al. Regulation of peripheral lymph node genesis by the tumor necrosisfactor family member TRANCE. J. Exp. Med. 192, 1467–1478 (2000).

10. Kong, Y. Y. et al. OPGL is a key regulator of osteoclastogenesis, lymphocytedevelopment and lymph-node organogenesis. Nature 397, 315–323 (1999).

11. Dougall, W. C. et al. RANK is essential for osteoclast and lymph nodedevelopment. Genes Dev. 13, 2412–2424 (1999).

12. Forster, R. et al. A putative chemokine receptor, BLR1, directs B cell migration todefined lymphoid organs and specific anatomic compartments of the spleen. Cell87, 1037–1047 (1996).

13. Ansel, K. M. et al. A chemokine-driven positive feedback loop organizes lymphoidfollicles. Nature 406, 309–314 (2000).

14. Banks, T. A. et al. Lymphotoxin-a-deficient mice. Effects on secondary lymphoidorgan development and humoral immune responsiveness. J. Immunol. 155,1685–1693 (1995).

15. De Togni, P. et al. Abnormal development of peripheral lymphoid organs in micedeficient in lymphotoxin. Science 264, 703–707 (1994).

16. Futterer, A., Mink, K., Luz, A., Kosco-Vilbois, M. H. & Pfeffer, K. The lymphotoxin breceptor controls organogenesis and affinity maturation in peripheral lymphoidtissues. Immunity 9, 59–70 (1998).

17. Koni, P. A. et al. Distinct roles in lymphoid organogenesis for lymphotoxins a and brevealed in lymphotoxin b-deficient mice. Immunity 6, 491–500 (1997).

18. Park, S. Y. et al. Developmental defects of lymphoid cells in Jak3 kinase-deficientmice. Immunity 3, 771–782 (1995).

19. Cao, X. et al. Defective lymphoid development in mice lacking expression of thecommon cytokine receptor gamma chain. Immunity 2, 223–238 (1995).

20. Coles, M. C. et al. Role of T and NK cells and IL7/IL7r interactions duringneonatal maturation of lymph nodes. Proc. Natl Acad. Sci. USA 103, 13457–13462(2006).

21. Hashi, H. et al. Compartmentalization of Peyer’s patch anlagen before lymphocyteentry. J. Immunol. 166, 3702–3709 (2001).

22. Rennert, P. D., Browning, J. L., Mebius, R., Mackay, F. & Hochman, P. S. Surfacelymphotoxin a/b complex is required for the development of peripheral lymphoidorgans. J. Exp. Med. 184, 1999–2006 (1996).

LETTERS NATURE | Vol 446 | 29 March 2007

550Nature ©2007 Publishing Group

23. Alimzhanov, M. B. et al. Abnormal development of secondary lymphoid tissues inlymphotoxin b-deficient mice. Proc. Natl Acad. Sci. USA 94, 9302–9307 (1997).

24. Jung, S. et al. In vivo depletion of CD11c1 dendritic cells abrogates priming of CD81

T cells by exogenous cell-associated antigens. Immunity 17, 211–220 (2002).25. de Graaff, E. et al. Differential activities of the RET tyrosine kinase receptor

isoforms during mammalian embryogenesis. Genes Dev. 15, 2433–2444 (2001).26. Pichel, J. G. et al. Defects in enteric innervation and kidney development in mice

lacking GDNF. Nature 382, 73–76 (1996).27. Moore, M. W. et al. Renal and neuronal abnormalities in mice lacking GDNF.

Nature 382, 76–79 (1996).28. Cacalano, G. et al. GFRa1 is an essential receptor component for GDNF in the

developing nervous system and kidney. Neuron 21, 53–62 (1998).29. Honma, Y. et al. Artemin is a vascular-derived neurotropic factor for developing

sympathetic neurons. Neuron 35, 267–282 (2002).30. Vargas-Leal, V. et al. Expression and function of glial cell line-derived neurotrophic

factor family ligands and their receptors on human immune cells. J. Immunol. 175,2301–2308 (2005).

31. de Boer, J. et al. Transgenic mice with hematopoietic and lymphoid specificexpression of Cre. Eur. J. Immunol. 33, 314–325 (2003).

32. Nishino, J. et al. GFR a3, a component of the artemin receptor, is required formigration and survival of the superior cervical ganglion. Neuron 23, 725–736(1999).

33. Peixoto, A., Monteiro, M., Rocha, B. & Veiga-Fernandes, H. Quantificationof multiple gene expression in individual cells. Genome Res. 14, 1938–1947(2004).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements The work described in this paper was funded by the MedicalResearch Council (MRC), UK. We thank C. Atkins and G. Preece for cell sorting;S. Pagakis, M. Tolaini, T. Norton and K. Williams for technical assistance. We alsothank H. Hamada and J. Nishino for the GFRa3 knockout mice. H.V.-F. and K.E.F.were supported by a grant from the European Union.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. The authors declare no competing financial interests.Correspondence and requests for materials should be addressed to D.K.(dkiouss@nimr.mrc.ac.uk).

NATURE | Vol 446 | 29 March 2007 LETTERS

551Nature ©2007 Publishing Group