Related pituitary cell lineages develop intointerdigitated 3D cell networksLionel Budrya,1, Chrystel Lafontb,c,d,e,1, Taoufik El Yandouzib,c,d,e, Norbert Chauvetb,c,d,e, Geneviève Conéjerof,Jacques Drouina,2, and Patrice Mollardb,c,d,e,2

aLaboratory of Molecular Genetics, Institut de Recherches Cliniques de Montréal, Montréal, QC, Canada H2W 1R7; bDepartment of Endocrinology, Instituteof Functional Genomics, cCentre National de la Recherche Scientifique Unité Mixte de Recherche 5203, and dInstitut National de la Santé et de la RechercheMédicale U661, 34094 Montpellier, France; eUniversité de Montpellier, 34094 Montpellier, France; and fInstitut National de la Recherche Agronomique,Unité Mixte de Recherche Biochimie et Physiologie Moléculaire des Plantes, 34060 Montpellier, France

Edited* by Wylie W. Vale, Salk Institute for Biological Studies, La Jolla, CA, and approved June 14, 2011 (received for review April 13, 2011)

The pituitary gland has long been considered to be a randompatchwork of hormone-producing cells. By using pituitary-scaletridimensional imaging for two of the least abundant cell lineages,the corticotropes and gonadotropes, we have now uncoveredhighly organized and interdigitated cell networks that reflecthomotypic and heterotypic interactions between cells. Althoughnewly differentiated corticotrope cells appear on the ventralsurface of the gland, they rapidly form homotypic strands of cellsthat extend from the lateral tips of the anterior pituitary along itsventral surface and into the medial gland. As the corticotropenetwork is established away from the microvasculature, cellmorphology changes from rounded, to polygonal, and finally tocells with long cytoplasmic processes or cytonemes that connectcorticotropes to the perivascular space. Gonadotropes differenti-ate later and are positioned in close proximity to corticotropes andcapillaries. Blockade of corticotrope terminal differentiation pro-duced by knockout of the gene encoding the transcription factorTpit results in smaller gonadotropes within an expanded cellnetwork, particularly in the lateral gland. Thus, pituitary-scaletridimensional imaging reveals highly structured cell networks ofunique topology for each pituitary lineage. The sequential de-velopment of interdigitated cell networks during organogenesisindicate that extensive cell:cell interactions lead to a highly orderedcell positioning rather than random patchwork.

reproduction | stress | systems biology | pro-opiomelanocortin |gonadotropins

Since its discovery and early characterization, the anterior pitu-itary gland has been described histologically as a patchwork of

cells with little evidence of higher-order organization. This tissue iscomposed in the adult of five secretory cell types, each dedicated toa production of a different hormone, together with support tissue inthe form of glial-like cells, the folliculostellate cells, and a capillarybed. All secretory lineages of the anterior pituitary, as well as theintermediate lobe (IL) melanotropes, have a common origin in theoral ectoderm. The IL is thus a much simpler tissue, as it containsonly the pro-opiomelanocortin (POMC)-expressing melanotropesthat process POMC into themelanotropic hormone α–melanocyte-stimulating hormone. With its five lineages, the developmentallyrelated anterior lobe (AL) is more complex: it also containsa POMC-expressing lineage, the corticotropes that process thesame POMC precursor into adrenocorticotropic hormone(ACTH), the gonadotrope cells that produce the gonadotropinsluteinizing hormone (LH) and FSH, the thyrotropes that produceTSH, the somatotropes that produce growth hormone (GH), andthe lactotropes that produceprolactin.TheALand ILare separatedby a cleft that is the remnant of the original lumen of the pituitaryprimordium, theRathke pouch. Interestingly, it was recently shownthat pituitary stem cells and progenitors are maintained along thatcleft in the adult pituitary (1, 2).The investigation of mechanisms for control of pituitary func-

tion led to the concept of “neuroendocrine” interactions (3) and

to the discovery of hypothalamic hypophysiotropic hormones thatare dedicated to the control of hormone secretion by pituitarycells (4, 5). The hypothalamus thus provides major inputs forcontrol of pituitary function, and this is balanced by feedbackregulation exerted by circulating hormones that inform the pitu-itary of the status of peripheral endocrine glands. In addition tothese critical interactions between central and peripheral inputs,a large body of evidence has supported the idea of autocrine andparacrine regulation within the pituitary: the structural basis forintrapituitary regulation has, however, remained unclear. An-other conundrum that remained poorly understood until recentlyis the significant difference in responsiveness observed whencomparing in vivo pituitary tissue with isolated pituitary cells. Therecent discovery of homotypic GH cell networks extendingthroughout the pituitary has provided the structural basis to ex-plain these types of interactions and differing responsiveness.Indeed, all GH cells of the pituitary maintain adherent junctionswith each other to form an extensive cell network that allowsextremely rapid exchange of signals between cells and the pro-duction of highly synchronized and high amplitude secretory re-sponse (6, 7). The existence of this 3D web-like GH cell networkwas not suspected before on the basis of traditional of 2D histo-logical analyses, particularly as GH cells account for the bulk ofanterior pituitary cells.We currently have little idea whether and how the less abun-

dant anterior pituitary cell lineages fit within this network andwhether they are themselves part of homotypic or heterotypic cellnetworks. The present work investigated the putative networkorganization of two less abundant anterior pituitary lineages, thecorticotropes and gonadotropes, and defined the relationshipbetween different cell networks andmicrovasculature. We furtherrevealed an ordered developmental process for establishment ofadult cell networks and showed how blockade of corticotroph dif-ferentiation affects cells and networks of the gonadotroph lineage.The interdigitated homotypic networks of cells of these two line-ages exhibit unique cell morphologies, positions relative to micro-vasculature and other cells, as well as overall tissue distributions.

ResultsHighly Structured Corticotrope Cell Network. To visualize the or-ganization of pituitary POMC cells, we used a line of POMC–EGFP transgenic mice that was previously characterized (8).

Author contributions: L.B., J.D., and P.M. designed research; L.B., C.L., T.E.Y., N.C., andG.C. performed research; L.B. and C.L. contributed new reagents/analytic tools; L.B., C.L.,T.E.Y., N.C., G.C., J.D., and P.M. analyzed data; and L.B., C.L., J.D., and P.M. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.1L.B. and C.L. contributed equally to this work.2To whom correspondence may be addressed. E-mail: jacques.drouin@ircm.qc.ca orpatrice.mollard@igf.cnrs.fr.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1105929108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1105929108 PNAS | July 26, 2011 | vol. 108 | no. 30 | 12515–12520

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

6, 2

020

These mice express GFP in pituitary POMC lineages and thedevelopmental pattern of EGFP expression follows closely thatof the endogenous POMC gene. The analysis of EGFP-positivecells was carried out by using two-photon excitation microscopyand fixed whole pituitaries. Classical coronal pituitary sectionsstained by immunohistochemistry for ACTH often present ALcorticotrope cells as isolated cells with discontinuous fragmentsof positive cytoplasm; we observed a similar distribution of GFPfluorescence in 5-μm sections (Fig. 1A). Tridimensional (i.e., 3D)reconstructions of similar sections from the same POMC–EGFPadult pituitaries revealed structured networks of closely associ-ated POMC cells (Fig. 1B). These reconstructions revealedstrands of POMC cells extending from the ventral surface towardthe middle of the AL, without reaching the dorsal side. The dorsalside of the gland exhibits the strongly GFP-positive IL cells to-gether with dispersed islets of similarly strong EGFP-positive cells(Fig. 1B). Throughout the AL, POMC cells appeared highlyinterconnected, suggestive of homotypic interactions (Fig. 1Cand Movie S1). These tight contacts between POMC cells areclearly visible within the strands of POMC–EGFP cells and also inlateral areas where cells are more dispersed. This assessment is

a likely underestimate because the EGFP transgene is not fullypenetrant (8).To situate the POMC cell network relative to the microcircu-

lation, we revealed endothelial cells by immunohistofluorescencelabeling of PECAM (CD31). These data showed that POMC–EGFP cell bodies are, for the most part, not juxtaposed tocapillaries (Fig. 1D). However, higher-resolution imaging of thesepreparations revealed numerous cytoplasmic projections orcytonemes that provide cell–cell contacts between differentPOMC cells and between POMC cells and blood vessels (Fig. 1E).Thus, AL POMC cells form homotypic networks and have directcontacts with the capillary bed.By contrast, the dorsal AL contains islets of EGFP-positive

polygonal cells with little or no cytonemes; these cells do notappear to be in contact with other AL POMC cells (Fig. 1F). Thestrength of the EGFP signal in these islets and their cell mor-phology are similar to that of IL melanotropes (Fig. 1G), sug-gesting that these cells may be of IL origin. To establish the identityof these dorsal POMC–EGFP islets, we used a melanotrope-specific marker, Pax7. Immunohistochemical analyses of Pax7costaining with POMC–EGFP clearly showed that Pax7 expres-sion is restricted to IL cells and to these dorsal cell islets and thatit is absent from AL corticotropes (Fig. 1H).

Developmental Dynamics of POMC Cell Networks. During de-velopment, the first POMC–EGFP cells appear as isolated cellson the ventral surface of the gland, suggesting that their differ-entiation and expansion occur stochastically (Fig. 2A). Indeed,the POMC–EGFP cells that appear at embryonic day (E) 13.5are small, rounded, and for the most part isolated from eachother (Fig. 2A); this is consistent with de novo differentiationrather than clonal expansion of a few differentiated cells (9).Quite rapidly after their initial detection, POMC–EGFP cellsappear to cluster together, and they are particularly abundant onthe caudal side of the ventral gland surface (Fig. 2B). Therounded cell appearance with clearly visible nuclei initially ob-served at E13.5 (Fig. 2A, Inset) is rapidly replaced by polygonalcells that exhibit less visible nuclei (E13.75; Fig. 2B, Inset). ByE14.5, the number of POMC–EGFP cells increases significantlyon the ventral surface, with the highest density on the caudal side(Fig. 2C). Viewed from the ventral surface, the lateral tips of thegland show POMC–EGFP cells organized in strands extendingfrom the ends. This organization is best seen at E15.5 (Fig. 2D).On the dorsal side, these strands come together at the lateralextremities but do not extend dorsally (Fig. 2E). In fact, very fewisolated EGFP-positive cells are present on the dorsal surface atthis stage. Between E15.5 and E17.5, the most striking differenceis a change in cell shape with the appearance of cytoplasmicprojections (Fig. 2 F and G). Coronal sections reveal that theventral strands of POMC–GFP cells penetrate into the AL at themediolateral level to reach the middle of the developing gland(Fig. 2H). This organization is conserved at P0 (Fig. 2I) andpostnatally into adulthood (Fig. 1B).

Interrelated LH Cell Network. As the POMC lineages were pro-posed to share a common precursor with gonadotropes (10), itwas particularly interesting to assess the organization of gona-dotropes and to compare it with that of corticotropes. For thispurpose, we generated LH-cerulean transgenic mice that usea bovine β-LH promoter to target expression of a ceruleanfluorescent transgene (11) in this lineage (12). Similar to the firstPOMC cells (Fig. 2A), the earliest (E17.5) LH-cerulean cellsappear in the midventral surface of the gland as isolated cells(Fig. 3A). By E18.5, LH-cerulean cells are clustered in the ven-tral mediolateral AL (Fig. 3B, Inset). These cells are closelypacked and always associated with strands of POMC–GFP cellsthat penetrate the AL (Fig. 3 B, D, and E). In addition, isolatedLH-cerulean cells present more laterally also have intimate

E

F

AL

PECAMPECAM

Pax7

H

AL

IL

C

G

50 µm

D

POMC-EGFP

Fig. 1. Tridimensional organization of POMC cells in adult mouse pituitary.(A) Distribution of POMC–EGFP fluorescence on 5-μm coronal sectionthrough an adult (P70) mouse pituitary revealing homogenously stained IL,negative posterior lobe (PL), and discontinuous fragments of fluorescentcytoplasm in AL. The x, y, and z axes are 960, 1,705, and 5 μm, respectively;inset is 100 × 100 × 5 μm. (B) Similar view of POMC–EGFP pituitary following3D reconstruction (depth, 70 μm) of images obtained by two-photon con-focal microscopy. Insets in A and B represent high-magnification views of thesame pituitary field in 2D and 3D, respectively. The x, y, and z axes are 960,1,705, and 70 μm, respectively; inset is 100 × 100 × 70 μm. (C) Higher-magnification 3D reconstruction of POMC–EGFP cell network viewed fromventral AL (Movie S1). The x, y, and z axes are 126, 126, and 57 μm, re-spectively. (D) Three-dimensional reconstruction of fluorescent coimaging ofPOMC–EGFP cells (green) with PECAM (CD31) labeling of endothelial cellsrevealing capillary bed of adult AL. The x, y, and z axes are 156, 100, and 20μm, respectively. (E) High-magnification 3D reconstruction of capillaries withPECAM (magenta) together with POMC–EGFP fluorescence. The x, y, and zaxes are 62, 46, and 16 μm, respectively. (F) Dorsal view of adult mouse pi-tuitary revealing islets of very intense EGFP-positive cells. The x, y, and z axesare 228, 181, and 94 μm, respectively. (G) High-resolution view of tightlypacked IL melanotrope cells positive for POMC–EGFP. (H) Colabeling of adultPOMC–EGFP pituitary for Pax7 (magenta) and POMC–EGFP. Coexpression(yellow) is observed in IL but not in AL (Inset). The x, y, and z axes are 204,134, and 4 μm; inset is 34 × 31 × 4 μm.

12516 | www.pnas.org/cgi/doi/10.1073/pnas.1105929108 Budry et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

6, 2

020

contact with POMC cells (Fig. 3C). This network is maintainedin the adult AL (Fig. 3 F–L). As for POMC cells, the LH cellsform homotypic networks within the AL (Fig. 3G and Movie S2).Colabeling with PECAM to reveal capillaries showed that LHcells are in close proximity to the pituitary microvasculature (Fig.3H), much closer in fact than POMC cells (Fig. 1F).The adult pituitary also revealed a dense population of LH-

cerulean cells on the dorsal AL surface (Fig. 3I); this streak ofLH cells extends under the IL, where very few POMC cells arepresent (Fig. 3I Inset). These dorsal LH cells appear during thefirst postnatal week [i.e., postnatal day (P) 2; Fig. 3J] and are fullypresent in the P15 pituitary (Fig. 3K), but not at E18.5 (Fig. 3D).It is noteworthy that these dorsal LH cells have almost no con-tact with POMC cells, as clearly visible on dorsal views of thegland (Fig. 3L).In summary, whereas the POMC cell network appeared to or-

ganize itself into strands emerging from the lateral extremitiesand the caudoventral surface of the gland, the LH cells mostlyappear/differentiate in themedioventral surface of the developinggland and develop between the POMC cell and capillary net-

works. In addition, another population of dorsal LH cells with fewcontacts with the POMC cell network appears postnatally.

Expanded LH Cell Network in Absence of Corticotropes. The associ-ation of LH cells with the POMC cell network, taken togetherwith their differentiation from a putative common precursor,suggests that development of the LH network may be dependent

E15.5

ventral view

P0

E17.5

ventral view

E15.5

dorsal view

POMC-EGFP

E13.5 E13.75

ventral viewventral view

E14.5

C

A B

R

C

R

C

R

C

E

E17.5E15.5 50 µm50 µm

F G

Fig. 2. Developmental dynamics of POMC cell morphology and networkorganization. (A) Ventral view of E13.5 POMC–EGFP mouse pituitary inwhich the first fluorescent signal appears as small and isolated cells (Inset). R,rostral; C, caudal. The x, y, and z axes are 540, 258, and 199 μm, respectively;inset is 62 × 62 × 199 μm. (B) By E13.75, POMC–EGFP–positive cells appear toregroup on the caudal side of the ventral surface (Inset). The x, y, and z axesare 916, 438, and 151 μm; inset is 62 × 62 × 151 μm. (C) At E14.5, denselypacked POMC–EGFP cells are seen on the caudal–ventral side of the AL. Thex, y, and z axes are 1,126, 584, and 205 μm, respectively. (D) At E15.5, strandsof POMC–EGFP cells extend from the lateral tip of the AL along its ventralsurface. The x, y, and z axes are 368, 368, and 107 μm. (E) These strands ofPOMC–EGFP cells come together at the lateral tip of the AL but do not ex-tend on the dorsal side of the gland. The x, y, and z axes are 368, 368, and 85μm, respectively. (F) POMC–EGFP cells of E15.5 pituitary have more polygo-nal appearance than at E13.5. (G) By E17.5, the POMC–EGFP cells extendcytoplasmic projections that maintain contacts between different POMC–EGFP cells. (H) Coronal section through E17.5 POMC–EGFP pituitary showsstrands of fluorescent cells extending from ventral surface of the AL into themediolateral sides of the gland. The x, y, and z axes are 980, 518, and 106μm, respectively. (I) A similar organization is present in the P0 postnatal pi-tuitary. The x, y, and z axes are 1,788, 453, and 103 μm, respectively.

POMC-EGFP LH-CerE17.5A

E18.5

B ventral view

E18.5

PECAMdorsal view

D

C

G

adult

H

IJ

adultP15P2

F

I

E

LK

E18.5

E18.5 adult

adult

50µm

Fig. 3. Development of LH cell network during pituitary organogenesis. (A)The first LH-cerulean (LH-Cer, magenta) cells appear at E17.5 as small cells onthe ventral surface of the AL. (B) Ventral view of double E18.5 transgenicpituitary for POMC–EGFP (green) and LH-cerulean show initial appearanceof LH-cerulean cells in the rostromedial AL, in contrast to POMC–EGFP cellsthat are more preeminent in the lateral wings. (R, rostral; C, caudal.) The x, y,and z axes are 1,374, 612, and 160 μm, respectively; inset is 124 × 124 × 160 μm.(C) View of lateral wing of E18.5 double transgenic pituitary revealing fewscattered LH-cerulean cells in intimate contact with strands of POMC–EGFPcells. The x, y, and z axes are 318, 318, and 160 μm, respectively. (D) Coronalsection through E18.5 double transgenic pituitary. Whereas POMC–EGFPcells are mostly present along the edges of the developing AL, theLH-cerulean cells extend from the medioventral surface into the ventral halfof the developing AL. The x, y, and z axes are 1,750, 613, and 102 μm, re-spectively. (E) High-magnification view of E18.5 double transgenic pituitaryshows intimate contacts between LH-cerulean cells and strands of POMC–EGFP cells penetrating the AL. The x, y, and z axes are 125, 77, and 102 μm,respectively. (F) Close contacts between LH-cerulean cells and strands ofPOMC–EGFP cells penetrating the AL are preserved in adulthood (P70 dou-ble transgenic male). The x, y, and z axes are 225, 225, and 41 μm, re-spectively. (G) Higher-magnification 3D reconstruction of LH-cerulean cellnetwork viewed from ventral AL (Movie S2). The x, y, and z axes are 125, 125,and 82 μm, respectively. (H) Colabeling for endothelial PECAM (CD31) andLH-cerulean reveals intimate contacts between gonadotropes and pituitarycapillaries (P70). The x, y, and z axes are 111, 106, and 30 μm, respectively. (I)Adult (P70) double transgenic male pituitary revealing the presence ofa dorsal population of LH-cerulean cells that form a continuous band ofgonadotropes along the dorsal side of the AL. Inset: These dorsal LH-cerulean cells have few contacts with POMC–EGFP cells. The x, y, and z axesare 1,902, 1,000, and 56 μm, respectively; inset is 225 × 225 × 46 μm. (J) At P2,the LH cell network remains concentrated along the medioventral AL andthe first LH cells are detected in the dorsal AL. The x, y, and z axes are 1,070,580, and 127 μm, respectively. (K) Three-dimensional reconstruction of LH-cerulean cells in AL of P15 pituitary reveals intimate contacts between dorsalgonadotropes. The x, y, and z axes are 833, 489, and 135 μm, respectively. (L)Dorsal view of adult (P70) double transgenic pituitary shows high density ofLH-cerulean cells that are contiguous with the homogenous POMC–EGFP–positive IL. The x, y, and z axes are 332, 360, and 100 μm, respectively.

Budry et al. PNAS | July 26, 2011 | vol. 108 | no. 30 | 12517

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

6, 2

020

on the POMC network. We assessed this hypothesis by in-vestigating the LH cells in Tpit−/− pituitaries that are deficientin POMC cell differentiation (10). Ventral views of Tpit−/−;LH-cerulean E18.5 pituitaries showed a marked increase in thenumber of LH cells, most significantly in lateral wings (Fig. 4B vs.Fig. 4A). These glands were also stained by IHC for the Tpit-lacZ chimera that is expressed in knockout cells that fail todifferentiate into either POMC cells (1–2% of cells) or switchinto gonadotropes (13). Accordingly, colabeled cells were notobserved and the decrease of Tpit-lacZ–positive cells in lateralwings of Tpit−/− pituitaries appeared compensated by LH-cerulean cells (Fig. 4B vs. Fig. 4A). Morphometric quantificationof LH cell mass strengthens the proposal that more LH cells arepresent in the lateral zones of Tpit−/− than Tpit+/− pituitaries(Fig. 4C). In the adult, the increase in LH cells is mostly visible inthe lateral wings (Fig. 4E vs. Fig. 4D). To quantitatively assesscell numbers, we performed FACS analysis of freshly dispersedpituitaries of both genotypes (Fig. 4F) and observed significantincreases in LH cell numbers (Fig. 4G). In addition, the LH cellsfrom Tpit−/− pituitaries were smaller by approximatey 30%compared with control siblings (Fig. 4 D and E, Insets, and Fig. 4F and H).

DiscussionBy using pituitary-scale 3D imaging during development (Fig.5A) and in adults (Fig. 5B), the present work revealed the unique3D topological organization of cell networks for two low-abundance pituitary lineages (each representing ≤10% of adultAL cells), the corticotropes and gonadotropes, that derive froma putative common precursor. These networks occupy different,yet interdigitated, positions within the adult gland. In particular,the position of the two networks relative to the microvasculatureis different: whereas gonadotropes are usually in close contactwith capillaries, the corticotropes compensate their distancefrom blood vessels by a different cytoarchitecture that includescytoplasmic projections or cytonemes. Cell shape and network

topology are thus adapted to provide equal access of cells tovasculature within the context of the other and more abundantlineages that occupy the bulk of the gland. Whereas cell net-works may serve an important role in endocrine function by therapid exchange of signals, as shown for the GH cell network (6, 7,14), their presence may also reflect the developmental program oforganogenesis (Fig. 5C). These two functions are not mutually ex-clusive and most likely reflect concerted developmental purposes.

Homotypic Cell Networks and Pituitary Cell Function. Contrary toprevious belief and patchwork appearance on tissue sections,anterior pituitary cells are organized in homotypic networks thatare readily visible on 3D imaging. Cell interactions within net-works are likely critical for concerted rapid responses (6, 14)and may contribute to adaptation to new endocrine conditionsin adult (7).Homotypic networks and cell shape are also adapted to pro-

vide equivalent access to the vasculature for quick cell responsesto circulating signals. Indeed, some cells have a cellular archi-tecture that allows them to be, broadly speaking, further awayfrom capillaries but nonetheless in contact through cytonemesthat directly reach the perivascular space. Interestingly, it is thecorticotropes, which are the first cells to differentiate duringorganogenesis, that adopt this cellular architecture and thusadapt their cell shape and homotypic network organization tothe appearance of other lineages.Close associations between cells and microvasculature ensure

(i) integration of hypothalamic inputs, which are gradually dis-tributed within the parenchyma; (ii) appropriate supplies of ox-ygen (and nutrients), which are finely adjusted depending on cellconsumption; and (iii) delivery of hormonal products toward theperivascular space, from which they will pass the endothelial cellbarrier into the bloodstream (14). Thus, although corticotropespossess somas that are distant from the capillaries, the genera-tion of ACTH secretion burst is not slower compared with otherhormones, as primed ACTH vesicles are oriented along the

LH-C

erTp

it-βg

al

% L

H-p

ostit

ive

mea

n ce

llvo

lum

e(µ

m3

X 10

3 )

+/- -/-+/+ +/- -/-+/+

R

C

ventral view

R

C

ventral view

E18.5 E18.5

lateral zone

+/- -/-

Cer

+ ce

ll m

ass

(µm

3X

106

0.5

0.3

0.2

0.1

0.0

0.4

LH-Cer X Tpit+/-

coun

ts

+/+-/-

35

8

0

17

26

0 64 128 192 256

LH-Cer X Tpit-/-

adult adult

5

3

2

1

0

4

14

12

10

8

6

4

2

0

6

A B

C

D E

F

G H*

*

*

LH-C

er

Fig. 4. Extended LH cell network, but decreased cell volume, in corticotrope-deficient Tpit−/− pituitary. (A and B) LH-cerulean–positive cells (magenta)present in rostromedial AL of control Tpit+/−;LH-cerulean pituitary extend laterally in Tpit−/−;LH-cerulean mice. Immunostaining for the Tpit–β-gal chimericprotein derived from the mutant Tpit allele (green) reveals cells destined for corticotrope lineage. (R, rostral; C, caudal.) The x, y, and z axes in A are 729, 516,and 67 μm, respectively; those in B are 719, 516, and 103 μm, respectively. (C) Morphometric quantitation of LH-cerulean–positive cells in the lateral quartersof the developing AL (mean ± SEM; n = 5). Statistical analysis (Mann–Whitney) indicates statistical difference (P = 0.009). (D and E) Adult LH-ceruleantransgenic pituitary reveals increased number of gonadotropes in AL of Tpit−/− compared with Tpit+/− mice. Insets: LH-cerulean–positive cells of Tpit−/− pi-tuitary may be smaller. The x, y, and z axes in D and E are 966, 680, and 105 μm, respectively; inset is 100 × 100 × 69 μm. (F) FACS analysis of LH-cerulean–positive cells from WT (+/+) and Tpit−/− pituitaries revealing smaller cell size in Tpit−/− pituitary (magenta). (G) Quantification by FACS of LH-cerulean–positivecells as percentage of sorted AL cells showing increased number of gonadotrope cells in Tpit−/− pituitaries (P = 0.02, Kruskal–Wallis one-way ANOVA; Tpit+/+,n = 3; Tpit+/−, n = 5; Tpit−/−, n = 5; mean ± SEM). (H) Cell volume assessed by FACS for adult LH-cerulean cells of the same Tpit genotypes as before showa statistically significant decrease of approximately 40% cell volume in Tpit−/− pituitaries (P = 0.03, Kruskal–Wallis one-way ANOVA).

12518 | www.pnas.org/cgi/doi/10.1073/pnas.1105929108 Budry et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

6, 2

020

processes that contact vessels (15, 16) and small secreted prod-ucts such as ACTH circulate rapidly (∼10 μm/sec) within theextracellular space (14).Close associations between homotypic cell networks may be

the basis for paracrine interactions. In the present work, thedeficit of corticotropes in Tpit−/− AL results in a significant de-crease of gonadotrope cell volume in parallel with increasedgonadotrope number. The later is likely a result of the alternatecell fate adopted by their common precursors (10). However, the30% to 40% decrease in gonadotrope cell volume of Tpit−/−

pituitaries (Fig. 4H) is most likely caused by the absence of sig-nals provided by corticotropes, as Tpit is not expressed ingonadotropes (10, 17). These signals could be secreted but aremore likely to be membrane-associated given the close contactsbetween cells of these two lineages. Paracrine interactions havebeen suggested previously to modulate hormone responsiveness

but never to involve control of cell volume (18). The networkorganization of pituitary lineages and their heterotypic inter-actions may thus serve an important maintenance function in theadult gland by coordinating the balance between lineages, andthus contribute to adjust cell number, volume, and function.Conversely, disruption of these networks may be implicated insubtle forms of hormone imbalance/disorders (19).

Ontogeny of Corticotrope and Gonadotrope Networks. Our resultsprovide support for a pituitary-scale model of temporally precisecontrol of positional determination of gonadotropes and corti-cotropes, two non–Pit-1–dependent cell lineages (Fig. 5C). It isnoteworthy that the first terminally differentiated cells of bothlineages appear as single cells on the ventral surface of the de-veloping gland (Fig. 5C, 1 and 5, respectively). For corticotropes,available markers, such as Tpit and POMC, appear within 12 h ofeach other, such that it is not possible to know whether the cellswere determined earlier. For gonadotropes, however, hormoneor LH-cerulean are detected from E16, but the earlier expressionof SF1 at E13.5 suggests that cells destined to become gonado-tropes may have been determined earlier (20), possibly con-comitant with corticotropes, in agreement with their origin froma common progenitor (10). It should be noted, however, thatgonadotropes may be heterogeneous and that the present workdocumented the appearance and organization of LH-expressingcells. As FSH-expressing gonadotropes do not completely over-lap with LH expression, in particular during development (21),their organization may differ.The early-born corticotropes represent likely candidates for

establishment of a 3D scaffold that serves to organize networksof other lineages, as development of the GH cell network doesnot commence until the E15.5/E16.5 transition (6) and the firstwave of gonadotropes appears in close proximity to corticotropesat approximately E17.5. Indeed, soon after their appearance,corticotropes regroup on the caudal–ventral side of the glandand at the tips of the lateral wings of the forming gland (Fig. 5C,2) to form strands of cells that then penetrate within the body ofthe AL (Fig. 5C, 3). It is unclear whether clustering of cortico-tropes arises from de novo differentiation or whether differen-tiated corticotropes migrate to associate in this fashion. TheE15.5 to E18.5 period of pituitary organogenesis is also charac-terized by a change in corticotrope shape and the formation ofextended cell processes (Fig. 5 A and B, Insets). This occurs atthe same time as the lateral expansion of cells from Pit-1–dependent lineages within the medial AL (22). In particular,appearance of GH cells within the body of the AL may con-tribute to the “stretching” of POMC cells that are already or-ganized in strands. POMC cell cytonemes may thus form ratherpassively as a result of this action and of strong homotypicinteractions; alternatively, an active process of cytoplasmic out-growth may lead to intercalation of POMC cells within the ex-tensive GH cell network (6).It is along the preformed corticotrope network that gonado-

tropes differentiate and cluster. In the midventral AL, gonado-tropes rapidly establish homotypic contacts but also heterotypiccontact with corticotropes (Fig. 5 B and C). In contrast, at theAL lateral ends, isolated LH-cerulean–positive gonadotropes aremore often observed in isolation from each other but in closecontact with the corticotrope network. Again, the apparent re-placement of corticotropes by gonadotropes in Tpit−/− pituitar-ies, particularly at the lateral ends of the AL, supports the modelthat they share a common precursor (10). The close associationof LH cells with the corticotrope cell network as they differen-tiate during organogenesis further suggests that cells of theselineages share an ability for cell–cell interactions that precedesterminal differentiation. This idea is supported by the recentcharacterization of primitive pituitary adenomas that sharemarkers of both lineages (23). Nonetheless, LH-cerulean cells do

1) first POMC cells (e13.5)2) lateral strands (e14.5)3) ventral strands (e15.5)

4) first melanotrophs (e15.5)

7) post-natal melanotroph islets

5) first LH cells (e17.5)6) post-natal LH cells

1

23

4

5

6

7

C developmental sequence

B adult

A embryo - e18.5 e13.5 e17.5

Fig. 5. Schematic representation of POMC (green) and LH (magenta) cellnetworks in developing E18.5 (A) and adult pituitary (B). Insets: Photographsof POMC (green) and LH (red) cells at indicated times of development toillustrate changes in cell shape. (C) Developmental sequence of appearance/changes in POMC and LH cell networks. 1, The first POMC cells appear onventral AL at approximately E13.5. 2, By E14.5, strands of POMC cells extendfrom lateral tips of the AL along the ventral surface. 3, The ventral strands ofPOMC cells penetrate the AL. 4, The first IL melanotropes appear at E15.5. 5,The first LH cells appear on ventral AL surface at E17.5 and populate medialAL. 6, Postnatal appearance of LH cell network on dorsal side of AL. Thesepostnatal LH cells have few contacts with POMC cells. 7, Postnatal appear-ance of POMC- and Pax7-positive melanotrope islets on dorsal surface of AL.

Budry et al. PNAS | July 26, 2011 | vol. 108 | no. 30 | 12519

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

6, 2

020

not adopt an organization similar to that of corticotropes in theTpit−/− AL, suggesting that the strand organization of cortico-tropes is an intrinsic property. Further, POMC cells exhibit themost important change in cell shape during organogenesis,starting from small round cells to become polygonal with longcytonemes (Fig. 5 A and B, Insets). Thus, corticotropes may havea cell-autonomous program to form a network that scaffoldsother cell networks.After birth, further positional determination takes place with

the appearance of gonadotropes along the Rathke cleft and thedorsal side of the AL (Fig. 5C, 6). This second wave of gona-dotrope differentiation yields sheets of homotypically connectedcells that rarely contact AL corticotropes or POMC/Pax7–positive cells (of presumed IL origin), which form sparsely dis-tributed clusters of polygonal cells on the dorsal AL surface (Fig.5C, 7). This arrangement of dorsal gonadotropes persists intoadulthood and may be related to recently proposed subgroups offunctionally distinct gonadotropes (21).

Perspective: Revisiting Pituitary Biology? The current view of pi-tuitary organogenesis was largely developed from spatiotemporalexpression patterns of transcription factors and signaling mole-cules using mostly midsagittal pituitary sections (24). The bulk ofprevious studies did not analyze postnatal pituitary cell positionand did not account for the fact that the medial portion of thepituitary gland only represents approximately 20% of adult pi-tuitary cells (6). The results of the present study thus demanda revised assessment of pituitary organogenesis because (i) cor-ticotropes and gonadotropes form large-scale cell networks thatmay function in a similar manner to the somatotrope network tointegrate and propagate cell responses; (ii) both Pit1-dependent(6) and Pit1-independent (i.e., in the present study) lineagesundergo developmental networking of terminally differentiatedendocrine cells, suggesting a generalized paradigm of cell net-working; (iii) a positional determination of 3D cell networks ispresent and may involve morphogen gradients and/or transcrip-tional factors from the lateral wings toward the medial parts ofthe gland; and (iv) cell networks interact during development, asevidenced by the dependence of gonadotropes on the presence

of a fully differentiated corticotrope scaffold. The interdigitatedcell network revealed in the present study in mice is consistentwith the more clustered pituitary cell organization observed inother species such as fish (25) in which cell–cell coupling wasshown (26). Homotypic cell networks may thus represent a dif-ferent evolutionary solution to a dependence on cell coupling forefficient hormone response. Further, interdigitated cell networksmay favor heterotypic interactions and effective coordination ofdifferent endocrine axes.In summary, this pituitary-scale 3D imaging study of the posi-

tional determination of corticotropes and gonadotropes acrossa lifespan has highlighted the importance of pituitary cell net-working in mouse pituitary development. The developmental pro-cesses that guide formation of endocrine cell networks clearly haveimplications for the proper functioning of the adult pituitary gland.

Materials and MethodsTpit mutant (10) and POMC–EGFP transgenic mice (8) were described pre-viously. Production of LH-cerulean transgenic mice is described in SI Materialsand Methods, and their characterization is described in Fig. S1. For imaging,tissues were fixed in situ by using solution PBS with 4% paraformaldehyde(PFA) and stored in PBS solution until image acquisition. Immunostainingwas performed as previously described (9). Animal experimentation was ap-proved by the animal ethics review committee of Institut de RecherchesCliniques de Montréal.

Multiphoton and confocal images of pituitaries were acquired as pre-viously reported (6) and are detailed in SI Materials and Methods. PituitaryLH-cerulean–positive cells were analyzed by FACS using a BD Biosciences cellsorter and Summit software (version 4.3; Beckman-Coulter).

ACKNOWLEDGMENTS. We are indebted to many colleagues for discussionsand suggestions on this work. We are grateful to Noriko Uetani for herartistic contribution to illustrations, to Éric Massicotte for expert assistanceand FACS analyses, and to Lise Laroche for secretarial support. This work wassupported by Canadian Institutes of Health Research grants (to J.D.); InstitutNational de la Santé et de la Recherche Médicale, Centre National de laRecherche Scientifique, Université Montpellier 1 and 2, Agence Nationalede la Recherche (Pit-Net Project), Institut Fédératif de Recherches 3, andRégion Languedoc Roussillon Grants (to P.M.); the Irish National Biophoton-ics and Imaging Platform; the Irish Government’s Programme for Research inThird-Level Institutions, Cycle 4; and Ireland’s European Union StructuralFunds Programmes 2007–2013.

1. Melmed S, Drouin J (2010) Pituitary development. The Pituitary, ed Melmed S (Aca-demic Press, Oxford), pp 3–19.

2. Fauquier T, Rizzoti K, Dattani M, Lovell-Badge R, Robinson IC (2008) SOX2-expressingprogenitor cells generate all of the major cell types in the adult mouse pituitarygland. Proc Natl Acad Sci USA 105:2907–2912.

3. Harris GW (1955) The function of the pituitary stalk. Bull Johns Hopkins Hosp 97:358–375.

4. Schally AV (1978) Aspects of hypothalamic regulation of the pituitary gland. Science202:18–28.

5. Guillemin R (1978) Peptides in the brain: The new endocrinology of the neuron. Sci-ence 202:390–402.

6. Bonnefont X, et al. (2005) Revealing the large-scale network organization of growthhormone-secreting cells. Proc Natl Acad Sci USA 102:16880–16885.

7. Sanchez-Cardenas C, et al. (2010) Pituitary growth hormone network responses aresexually dimorphic and regulated by gonadal steroids in adulthood. Proc Natl Acad SciUSA 107:21878–21883.

8. Lavoie PL, Budry L, Balsalobre A, Drouin J (2008) Developmental dependence onNurRE and EboxNeuro for expression of pituitary proopiomelanocortin. Mol Endo-crinol 22:1647–1657.

9. Bilodeau S, Roussel-Gervais A, Drouin J (2009) Distinct developmental roles of cellcycle inhibitors p57Kip2 and p27Kip1 distinguish pituitary progenitor cell cycle exitfrom cell cycle reentry of differentiated cells. Mol Cell Biol 29:1895–1908.

10. Pulichino AM, et al. (2003) Tpit determines alternate fates during pituitary cell dif-ferentiation. Genes Dev 17:738–747.

11. Rizzo MA, Springer GH, Granada B, Piston DW (2004) An improved cyan fluorescentprotein variant useful for FRET. Nat Biotechnol 22:445–449.

12. Quirk CC, Lozada KL, Keri RA, Nilson JH (2001) A single Pitx1 binding site is essentialfor activity of the LHbeta promoter in transgenic mice. Mol Endocrinol 15:734–746.

13. Bilodeau S, et al. (2006) Role of Brg1 and HDAC2 in GR trans-repression of the pi-tuitary POMC gene and misexpression in Cushing disease. Genes Dev 20:2871–2886.

14. Lafont C, et al. (2010) Cellular in vivo imaging reveals coordinated regulation of pituitarymicrocirculation and GH cell network function. Proc Natl Acad Sci USA 107:4465–4470.

15. Moriarty GC (1973) Adenohypophysis: Ultrastructural cytochemistry. A review. J His-

tochem Cytochem 21:855–894.16. Moriarty GC, Halmi NS, Moriarty M (1975) The effect of stress on the cytology and

immunocytochemistry of pars intermedia cells in the rat pituitary. Endocrinology 96:

1426–1436.17. Lamolet B, et al. (2001) A pituitary cell-restricted T box factor, Tpit, activates POMC

transcription in cooperation with Pitx homeoproteins. Cell 104:849–859.18. Denef C (2008) Paracrinicity: The story of 30 years of cellular pituitary crosstalk. J

Neuroendocrinol 20:1–70.19. Waite E, et al. (2010) Different degrees of somatotroph ablation compromise pitui-

tary growth hormone cell network structure and other pituitary endocrine cell types.

Endocrinology 151:234–243.20. Ingraham HA, et al. (1994) The nuclear receptor steroidogenic factor 1 acts at multiple

levels of the reproductive axis. Genes Dev 8:2302–2312.21. Wen S, Ai W, Alim Z, Boehm U (2010) Embryonic gonadotropin-releasing hormone

signaling is necessary for maturation of the male reproductive axis. Proc Natl Acad Sci

USA 107:16372–16377.22. Ward RD, Stone BM, Raetzman LT, Camper SA (2006) Cell proliferation and vascu-

larization in mouse models of pituitary hormone deficiency. Mol Endocrinol 20:

1378–1390.23. Suzuki M, et al. (2008) ACTH and alpha-subunit are co-expressed in rare human pi-

tuitary corticotroph cell adenomas proposed to originate from ACTH-committed early

pituitary progenitor cells. Endocr Pathol 19:17–26.24. Zhu X, Gleiberman AS, Rosenfeld MG (2007) Molecular physiology of pituitary de-

velopment: Signaling and transcriptional networks. Physiol Rev 87:933–963.25. Liu NA, et al. (2008) In vivo time-lapse imaging delineates the zebrafish pituitary

proopiomelanocortin lineage boundary regulated by FGF3 signal. Dev Biol 319:

192–200.26. Levavi-Sivan B, Bloch CL, Gutnick MJ, Fleidervish IA (2005) Electrotonic coupling in the

anterior pituitary of a teleost fish. Endocrinology 146:1048–1052.

12520 | www.pnas.org/cgi/doi/10.1073/pnas.1105929108 Budry et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

6, 2

020