Thymospheres Are Formed by Mesenchymal Cells with the Potential to Generate Adipocytes, but Not Epithelial Cells

Report

Thymospheres Are Formed by Mesenchymal Cells with the Potential to Generate Adipocytes, but Not Epithelial Cells Graphical Abstract

Authors Julie M. Sheridan, Ashleigh Keown, Antonia Policheni, ..., Jakub Abramson, Tracy S.P. Heng, Daniel H.D. Gray

Correspondence [email protected]

In Brief The phenotype of putative bipotent epithelial stem/progenitor cells in the adult thymus has been controversial. Sheridan et al. find that a prominent candidate, the FoxN1 EpCAM sphereforming cell, harbors no epithelial stem/ progenitor capacity and present evidence that they are a unique intrathymic mesenchymal stromal/stem cell population with adipocyte differentiation capacity.

Highlights d

Thymosphere-forming cells (TSFCs) and fibroblasts share an origin and phenotype

d

TSFCs have adipocyte, but not thymic epithelial cell, differentiation capacity

d

TSFCs may be intrathymic mesenchymal stromal/stem cells of the adult thymus

Sheridan et al., 2017, Cell Reports 21, 934–942 October 24, 2017 ª 2017 https://doi.org/10.1016/j.celrep.2017.09.090

Cell Reports

Report Thymospheres Are Formed by Mesenchymal Cells with the Potential to Generate Adipocytes, but Not Epithelial Cells Julie M. Sheridan,1,2 Ashleigh Keown,1 Antonia Policheni,1,2 Siti N.A. Roesley,3 Noa Rivlin,4 Noam Kadouri,4 Matthew E. Ritchie,1,2 Reema Jain,1,2 Jakub Abramson,4 Tracy S.P. Heng,3 and Daniel H.D. Gray1,2,5,* 1Walter

and Eliza Hall Institute of Medical Research, Parkville, VIC, Australia of Medical Biology, University of Melbourne, Melbourne, VIC, Australia 3Monash Biomedicine Discovery Institute, Department of Anatomy and Developmental Biology, Monash University, Clayton, VIC, Australia 4Department of Immunology, Weizmann Institute of Science, Rehovot 76100, Israel 5Lead Contact *Correspondence: [email protected] https://doi.org/10.1016/j.celrep.2017.09.090 2Department

SUMMARY

Evidence suggests that a stem-cell-driven differentiation hierarchy maintains the dynamic thymic epithelial cell (TEC) network that governs T lymphocyte development. The identification of TEC stem/progenitor cells has been a major focus in the field, and several candidates with contrasting phenotypes have been described. We sought to determine the provenance and function of the only population reported to exhibit TEC stem cell properties in the adult, a Foxn1 EpCAM cell that generates socalled thymospheres. We provide evidence that the thymosphere-forming cell (TSFC) is not a TEC stem cell but can incorporate bystander TECs into thymospheres, providing an explanation for the epithelial activity ascribed to these structures. TSFCs were found to share a phenotype, transcriptional profile, and developmental origin with thymic fibroblasts and can generate adipocytes. In summary, this study redefines the nature of bipotent TEC stem/progenitor cells in the adult thymus and highlights a potentially important mesenchymal progenitor population. INTRODUCTION The structural and functional foundation of the thymus is a reticular network of diverse thymic epithelial cells (TECs) (Boyd et al., 1993). Together with other stromal elements, such as mesenchymal fibroblasts, endothelial cells, and dendritic cells, TECs of the outer cortex (cTECs) and inner medullary regions (mTECs) orchestrate T lymphocyte differentiation (Boyd et al., 1993; Pet€cker, 2007). However, at a relatively young rie and Zu´n˜iga-Pflu age the thymus undergoes progressive involution, resulting in impaired naive T cell production that is associated with diminished adaptive immune response in later life (Chaudhry et al., 2016). The involuted thymus is marked by alterations in epithelial cell composition and function, as well as fibrosis and adipogenesis (Chaudhry et al., 2016). Understanding how the stromal cell

types interrelate and their role in the maintenance, loss, and regeneration of the thymus is an important goal toward correcting these clinically important processes. It has been hypothesized that a stem cell-based hierarchy maintains the thymic epithelium. Several lines of evidence support the existence of an adult TEC stem cell, and the characterization of the putative stem cell and the lineage relationships that might support thymic function has been a major focus in the field (Boehm and Swann, 2013). Retrospective clonal analyses have demonstrated the existence of postnatal bipotent thymic epithelial progenitor cells (TEPCs) and lineage-restricted progenitors of unknown phenotype in the postnatal thymus (Bleul et al., 2006; Rodewald et al., 2001). However, attempts to prospectively identify adult TEPC populations analogous to those observed in the fetal thymus have met with limited success (Mayer et al., 2016; Ohigashi et al., 2015; Sekai et al., 2014). Similarly, while lineage tracing studies have demonstrated an ontological relationship between fetal TEPCs and major mTEC and cTEC populations in the adult thymus, postnatal lineage tracing has provided little insight beyond the neonatal period (Mayer et al., 2016; Ohigashi et al., 2013, 2015). Recently, several groups have prospectively identified populations that harbor bipotent differentiation capacity (Ucar et al., 2014; Ulyanchenko et al., 2016; Wong et al., 2014). The greatest resolution has been achieved using an EpCAM+ LY51posMHCIIhiPLET-1+ phenotype to identify a rare Foxn1-expressing population capable of long-term contribution to grafted thymic organoids, perhaps involving an EpCAM+LY51+MHCIIlo intermediary (Ulyanchenko et al., 2016; Wong et al., 2014). However, no FOXN1+EpCAM+ cells have yet been definitively shown to exhibit the cardinal stem cell characteristics of both differentiation and self-renewal. Intriguingly, a postnatal FOXN1 lineage cell has been reported to exhibit stem cell traits (Ucar et al., 2014). This was unexpected, since previous studies had demonstrated that TECs arise from a domain of the third pharyngeal pouch endoderm that expresses Foxn1 (Corbeaux et al., 2010; Gordon et al., 2001, 2004; Le Douarin and Jotereau, 1975). Using a strategy pioneered in other tissues to enrich stem/progenitors cells, Foxn1 EpCAM MHCII SCA-1+ cells were found to generate non-adherent colonies, termed thymospheres, when cultured

934 Cell Reports 21, 934–942, October 24, 2017 ª 2017 This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

A

B

Lin-

C

Lin- EpCAM68.7

EpCAM

MTS-15

52.5

38.7

7.8

SCA-1

SCA-1 E

Frequency (%)

D

14.6

Cell Dose

SCA-1+ MTS-15-

SCA-1+ MTS-15+

10,000

3/3

6/6

5,000

5/5

4/4

2,000

1/1

-

1,000

3/8

5/10

930

0/1

-

517

-

0/1

TSFC Frequency

1/1715 (1/805-1/3656)

1/1384 (1/653-1/2934)

under certain conditions (Ucar et al., 2014). With an apparent capacity to self-renew ex vivo, thymospheres were found to contribute to mature TEC subsets when placed in thymic organoids and grafted into mice (Ucar et al., 2014). This finding suggested that the TSFC could be the foundational unit of a postnatal TEC differentiation hierarchy, which is dependent upon initiation of FOXN1 expression for differentiation to functional maturity. In this study, we sought to refine the identity of this candidate TEC stem cell and define its relationship to FOXN1+ TEPCs and mature TECs. Our data challenge the notion that TSFCs are TEC stem cells. Instead, we find that TSFCs are non-epithelial stromal cells that share a phenotype and ancestry with thymic fibroblasts, and resemble mesenchymal stem cells in their expression profile and differentiation capacity. These findings redefine the postnatal TEC differentiation hierarchy and clarify the origin, composition, and differentiation capacity of the non-epithelial stromal cell compartment.

Figure 1. Two Phenotypically Distinct TSFC Populations in the Adult Thymus (A) Flow cytometric analysis of EpCAM versus SCA-1 staining on CD45 TER-119 CD31 (Lin ) thymus cells (representative of 10 experiments). See also Figure S1. (B) Representative image of thymosphere culture derived from Lin EpCAM SCA-1+ cells plated at 15,000 cells/mL. Inset: magnified region showing thymospheres with typical morphology. Scale bar, 100 mm. (C) Flow cytometric analysis of Lin EpCAM thymus cells showing SCA-1 versus MTS-15 staining (representative of 7 experiments). (D) Frequencies of MTS-15+ cells and MTS-15 cells as a percentage of total Lin EpCAM SCA-1+ population (mean ± SEM, n = 7 experiments). (E) Limiting dilution analysis of Lin EpCAM SCA-1+ subpopulations. Indicated numbers of cells were plated in thymosphere cultures and maintained for 7 days. Numbers of thymosphere-containing wells at each cell dose was used to calculate TSFC frequency. Estimates of frequency ± 95% confidence intervals are given; data are from 4 experiments (p = 0.703).

RESULTS

enrich for sphere-forming potential by excluding fibroblasts from our analysis; these are prevalent in the EpCAM population but do not share an ancestry with TECs (Foster et al., 2008; Gordon et al., 2004; Le Douarin and Jotereau, 1975; Le Lie`vre and Le Douarin, 1975). Flow cytometric analysis with the fibroblast marker MTS-15 (Gray et al., 2007) revealed that two populations comprise the EpCAM SCA-1+ thymosphere-forming population: a large population of MTS-15+ presumptive thymic fibroblasts (72.8% ± 3.8) and a smaller MTS-15 population (27% ± 3.7%) of unknown identity (Figures 1C and 1D). We hypothesized that excluding the MTS-15+ fibroblast population would result in greater than 3-fold enrichment of TSFCs. Surprisingly, we observed no significant difference in sphere-forming efficiency between these populations (p = 0.703); thus, depletion using MTS-15 did not enrich TSFCs (Figure 1E). Based on population size and TSFC frequencies, we deduced that the MTS-15+ population, previously characterized as being fibroblastic and keratin , contained the majority (80.5%) of TSFCs (Figures 1D and 1E) (Gray et al., 2007).

Adult EpCAM TSFCs Share a Phenotype with Thymic Fibroblasts To elucidate the TEC progenitor hierarchy in the postnatal thymus, we sought to better define what was presumably the most proximal of these, the recently identified Foxn1 TSFC. As a starting point, we fractionated adult thymic stromal cells using expression of the cell-surface proteins EpCAM and SCA-1 and tested for thymosphere-forming capacity under conditions previously described (Figures 1A and S1A) (Ucar et al., 2014). As reported, thymosphere generation was limited to the CD45 CD31 TER-119 (Lin ) EpCAM SCA-1+ stromal fraction (Figure 1B; data not shown) (Ucar et al., 2014). We next sought to

TSFCs Are Not Thymic Epithelial Stem Cells Thymospheres have been reported to stain with anti-keratin antibodies (Ucar et al., 2014), a definitive feature of epithelial cells (Moll et al., 2008). However, since MTS-15 had previously been described to label non-epithelial cells and we found that both MTS-15+ and MTS-15 cells generated thymospheres, we asked how keratin expression might arise in these structures. We generated spheres from purified Lin EpCAM SCA-1+ cells to confirm that they contained keratin+ cells by immunofluorescent detection. In contrast to previously published data on thymospheres that were generated from unfractionated CD45 stromal cells (Ucar et al., 2014), using the same broad-spectrum

Cell Reports 21, 934–942, October 24, 2017 935

A

B

D

C

E

F

anti-keratin antibody, we found few (if any) keratin+ epithelial cells in thymospheres (Figure 2A). To determine whether keratin+ TSFCs might extinguish keratin expression in culture, we looked for evidence of epithelial cells within the TSFC populations using flow cytometric analysis. Contrary to the data presented by Ucar et al., we found that few Lin EpCAM cells expressed keratins (1.7% ± 0.33%) (Figures 2B, 2C, and S2). Our flow cytometry data were confirmed by immunofluorescent analysis of keratin expression in purified thymic cell populations (2.5% ± 1.1% of Lin EpCAM cells) (Figures 2C and 2D), and both detection methods showed that keratin+ cells were almost entirely confined to the Lin EpCAM+ population (Figure 2D). Although these data are contrary to the findings of Ucar et al. (which showed that approximately one-

936 Cell Reports 21, 934–942, October 24, 2017

Figure 2. TSFCs and Their Thymosphere Progeny Are Not Epithelial (A) Maximal intensity projection immunofluorescence image of a thymosphere generated from Lin EpCAM SCA-1+ cells showing anti-keratin staining (red) counterstained with DAPI (nuclei, blue). Scale bar, 25 mm. Data are representative of 6 independent experiments. (B) Representative flow cytometric analysis of Lin thymus cells showing EpCAM versus keratin expression. Numbers indicate the percentage of cells within the respective regions and are representative of greater than 3 samples in 3 independent experiments. See also Figure S2. (C) Representative immunofluorescence images of flow cytometrically purified Lin EpCAM+, Lin EpCAM MTS-15+, and Lin EpCAM MTS-15 subpopulations stained with anti-keratin antibodies (green) and DAPI (nuclei, red). Scale bars, 40 mm. Data are representative of 3 independent experiments. (D) Frequency (%) of keratin+ cells in individual populations as determined by flow cytometry (black) and immunofluorescence (white) (mean ± SEM; representative of 3 independent experiments). (E) Flow cytometric analysis of a grafted RTOC generated from E15.5 BALB/c thymus cells (H-2Kb ) spiked with or without thymospheres derived from EpCAM YFP cells (H-2Kb+) from Foxn1cre;R26R-EYFP mice. Identification of thymosphere-derived cells in the Lin compartments using anti-H-2Kb antibody in thymosphere-containing (left, center) and carrier-only graft (right). (F) Comparison of YFP expression induced by Foxn1cre activity, and EpCAM versus MHC class II expression on carrier (H-2Kb ) and thymocytederived (H-2Kb+) cells in a graft (identified with an asterisk [*] in Table S1).

third of EpCAM cells were keratin+), they are consistent with several other studies that found that keratin expression is almost exclusive to the EpCAM+MHCII+ compartment (Gray et al., 2008; Rossi et al., 2007). These data indicated that TSFCs were not epithelial. To test the possibility that TSFCs could contribute to mature thymic epithelial subtypes following culture, we turned to a reaggregation thymic organ culture and grafting system. We crossed mice expressing Cre recombinase under the control of Foxn1 gene promoter (Foxn1Cre) (Zuklys et al., 2009) with those expressing a Cre-inducible YFP reporter (R26R-EYFP mice) (Srinivas et al., 2001) to indelibly label differentiated TECs with YFP. As expected, thymospheres from these mice were YFP , since TSFCs and their immediate progeny do not express Foxn1 (Ucar et al., 2014). Foxn1YFP thymospheres (H-2Kb) were reaggregated with major histocompatibility complex (MHC) mismatched BALB/c (H-2Kd) fetal thymus carrier cells then grafted under the renal capsule of recipient mice. After 4–10 weeks, thymic grafts were analyzed by flow cytometry to determine

A

B

Figure 3. EpCAM EpCAM+ TECs

C

Cell-Derived Thymospheres Can Incorporate

(A) Maximal intensity projection image of immunofluorescent staining of thymospheres generated from purified Lin EpCAM cells (left) and unfractionated CD45 cells (right) showing nuclei (DAPI, blue) and keratin (red). Data are representative of 6 experiments. Scale bars, 20 mm. (B) Quantification of keratin+ bodies per thymosphere. 18–24 z stacks were scored per group from 5–6 experiments. Groups were compared using an unpaired t test with Welch’s correction (****p = < 0.0001). (C) Maximal intensity projection of immunofluorescent staining of a thymosphere generated from wild-type Lin EpCAM cells cultured in the presence of YFP+EpCAM+ TECs (DAPI [nuclei], blue; anti-YFP, green). Scale bar, 25 mm.

whether H-2Kb thymosphere-derived cells had contributed to MHCII+ thymic epithelium. Two out of thirteen grafts showed evidence of thymosphere contribution above the background levels exhibited by grafts containing only carrier cells (Figure 2E; Table S1). In these grafts, the contribution of thymospherederived cells remained below 2% of total Lin stromal cells, despite inclusion of thymospheres generated from the equivalent of 3.5–4 mice (Figure 2E; Table S1). Foxn1Cre-mediated YFP expression was not observed in thymosphere-derived cells, indicating that TSFCs were not capable of generating Foxn1+ progeny (Figure 2E; Table S1). Importantly, we found no evidence of thymosphere differentiation into functionally competent MHCII+EpCAM+ TEC compartments (Figure 2F; Table S1). EpCAM Thymospheres Can Incorporate ‘‘Passenger’’ TECs Since the thymospheres containing keratin+ cells generated by Ucar et al. were formed from unfractionated CD45 stromal cells (which include EpCAM+ TECs) (Ucar et al., 2014), we reasoned that those keratin+ cells could have been formed under the influence of cells outside of the EpCAM compartment. For instance,

EpCAM cells may be induced to express keratin by other cell types in mixed cultures. Alternatively, while it is clear that conventional EpCAM+ TECs are unable to generate thymospheres de novo, they might be incorporated into thymospheres generated by EpCAM cells (Ucar et al., 2014). To investigate these possibilities, thymic cells were depleted of CD45+ hematopoietic cells and either directly placed in thymosphere culture as total CD45 stroma or subjected to fluorescence-activated cell sorting (FACS) purification of Lin EpCAM cells and then established in parallel cultures. Immunofluorescence analysis revealed that in contrast to thymospheres generated solely from Lin EpCAM cells, those generated from CD45 stromal cells showed substantial keratin expression (Figures 3A and 3B). To determine whether keratin+ cells were TECs incorporated into the thymospheres, we analyzed thymospheres derived from 1:1.5 mixtures of wild-type EpCAM stromal cells and YFP+EpCAM+ TECs (this ratio reflects the proportions in the unfractionated CD45 population). YFP+ cells were observed in resultant thymospheres, confirming that conventional TECs had been incorporated as ‘‘bystander’’ cells (Figure 3C). TSFCs Are Mesenchymal Stromal Cells Since TSFCs exhibited properties incompatible with an epithelial identity, we next sought to determine their true nature using lineage tracing and gene expression profiling. Previous studies have shown that TECs are derived from the endoderm, while thymic fibroblasts are derived from the mesoderm and neural crest cells (NCCs) (Foster et al., 2008; Gordon et al., 2004; Komada et al., 2012; Le Douarin and Jotereau, 1975). To investigate the origin of TSFCs, we crossed mice expressing the Cre recombinase under the control of different gene promoters with the R26REYFP line to indelibly label (1) endoderm-derived TECs (using Foxn1Cre; Zuklys et al., 2009), (2) ectodermal NCC-derived cells (using Wnt1Cre; Danielian et al., 1998), or (3) mesoderm-derived cells (Mesp1Cre; Saga et al., 1999). As expected, YFP+ cells were confined to the EpCAM+ compartment in Foxn1Cre;R26REYFP mice, while the frequency of Wnt1cre YFP+ and MesP1cre YFP+ cells in the EpCAM+ fraction was not above background levels (p = 0.93 and p = 0.66, respectively) (Figures 4A and 4B). NCC-derived cells constituted approximately half (57.8% ± 7.2%) of the TSFC population, whereas mesoderm-derived cells were a minor component (9.5% ± 2.3%) (Figures 4A and 4B). Interestingly, most (82.5% ± 0.85%) MTS-15+ cells were derived from NCCs (Figures 4A and 4B) and given the limits of Credependent recombination, it remains possible that all MTS-15+ cells are NCC derived. In contrast, almost all mesoderm-derived cells were MTS-15 and constituted approximately one-quarter of cells (25.7 ± 3.5%) in this compartment, which is comparable to the NCC contribution (18.9% ± 3.1%) (Figures 4A and 4B). These data raise the possibility that two distinct populations of TSFCs exist within the postnatal thymus. Notably, purified Lin EpCAM SCA-1+YFP+ cells from Wnt1cre;R26R-EYFP mice readily generated YFP+ thymospheres (Figure 4C). This finding confirms that the majority of TSFCs are NCC derived and do not share an endodermal origin with conventional EpCAM+ TECs. Accumulating evidence suggests that thymic fibroblasts play an important role in the maintenance of postnatal thymic

Cell Reports 21, 934–942, October 24, 2017 937

A

B

C

D

Figure 4. TSFCs Are Mesenchymal Cells with Adipocyte Differentiation Capacity (A) Flow cytometric analysis of adult thymus cells from Foxn1cre;R26R-EYFP (top), Wnt1cre;R26R-EYFP (middle) and MesP1cre;R26R-EYFP (bottom) mice showing YFP expression in Lin EpCAM+ and Lin EpCAM SCA-1+ subpopulations. Profiles are representative of 5 individual mice from two experiments. (B) Frequency (%) of YFP+ cells in Lin EpCAM+, Lin EpCAM SCA-1+, Lin EpCAM SCA-1+MTS-15+, and Lin EpCAM SCA-1+MTS-15+ populations for indicated Cre recombinase mouse lines crossed with the R26R-EYFP mouse line (mean ± SEM; n = 3–8 mice from 2–4 independent experiments). Control group composed of Foxn1cre+/+; R26RYFP/+, Wnt1cre+/+; R26RYFP/+, and MesP1cre+/+; R26RYFP/+ mice. Lin EpCAM+: control versus Foxn1cre, p = < 0.0001; control versus Wnt1cre, p = 0.48; control versus MesP1cre, p = 0.99. Lin EpCAM SCA-1+: control versus Foxn1cre, p = 0.1; control versus Wnt1cre, p = 0.001; control versus MesP1cre, p = 0.017. Groups were compared using an unpaired t test with Welch’s correction (*p % 0.05; **p = % 0.01; ****p % 0.0001).

(legend continued on next page)

938 Cell Reports 21, 934–942, October 24, 2017

architecture and function (Lax et al., 2012; Sun et al., 2015); however, we have little understanding of how this compartment is maintained. Interestingly, a rare population with the capacity to retain a DNA label was found to have an overlapping phenotype with TSFCs (Osada et al., 2013; Ucar et al., 2014). Cell lines expanded from these cells had an MSC-like capacity to generate osteoblasts, adipocytes, and chondrocytes (Osada et al., 2013). These data support the hypothesis that a thymic mesenchymal stromal/stem cell (MSC) population maintains non-epithelial stromal populations, including fibroblasts and adipocytes. To investigate the possibility that TSFCs might fulfill this role, we compared gene expression profiles of TECs, thymospheres, and MSCs. We found that thymospheres shared a gene expression signature with mesenchymal cells (including Pdgfra, Pdgfrb, extracellular matrix [ECM] components, and growth factors), but not those characteristic of TECs (E-cadherin, occludin, keratins, MHC class II, and delta-like 4), and overall were more similar to MSCs than TECs (Figures S3A and S3B). Moreover, we observed robust adipocyte differentiation from TSFCs, suggesting that TSFCs are capable of giving rise to both thymic fibroblasts and adipocytes in the postnatal thymus (Figure 4D). DISCUSSION A postnatal TSFC of the Foxn1 lineage has been proposed as a candidate for the elusive TEC stem cell (Ucar et al., 2014). The phenotype of this putative TEC stem cell was consistent with data showing that FOXN1 is not required for initial specification events but is necessary for the subsequent differentiation to mature subsets (Nowell et al., 2011). TECs with experimentally reduced FOXN1 levels can retain thymic epithelial identity and persist for at least 6 months as ‘‘trapped’’ progenitors (Jin et al., 2014). The demonstration that Foxn1 TSFCs seemingly contributed to mature FOXN1+ TEC subsets implied that TSFCs were more primitive than all other adult TEPC populations described. We could readily form thymospheres using the strategies reported by Ucar et al. and confirm that thymospheres are derived only from the Lin EpCAM SCA-1+ compartment. However, our experiments indicate that TSFCs are not TEC stem cells but rather non-epithelial stromal cells that share an ancestry and transcriptional profile with thymic fibroblasts. Although it remains formally possible that a FOXN1 TEC stem cell exists in the EpCAM fraction, our results demonstrate that the TSFC does not fit this description. In light of these new data, the parsimonious hypothesis is that TEC regeneration capacity lies in the EpCAM+ compartment, since this is the only population that continues to demonstrate TEC differentiation capacity in the adult thymus (Ulyanchenko et al., 2016; Wong et al., 2014). Our finding that TSFCs do not differentiate into TECs is at odds with a previous report (Ucar et al., 2014). In that study, individual thymospheres generated from unfractionated CD45 thymic

stromal cells plated at high densities expressed both mTEC-specific and cTEC-specific keratin types. In conjunction with evidence suggesting that thymospheres could be clonally derived at low plating densities, the authors concluded that these findings together provided evidence of stem cell differentiation (Ucar et al., 2014). We have shown that thymospheres from cells plated at higher densities, such as those used for phenotyping and functional testing by Ucar et al., typically form polyclonally. This finding is directly analogous to data from spheres in other organ systems that are prone to aggregation yet form clonally at extremely low densities (Pastrana et al., 2011). Importantly, these data undermine the conclusion that a single EpCAM TSFC is necessarily the ancestor of the diverse keratin+ TECs observed in thymospheres. Our demonstration that EpCAM+ TECs can contribute as passengers to spheres generated by EpCAM TSFCs is sufficient to explain the appearance of keratin+ cells in thymospheres generated from populations that contain TECs (Ucar et al., 2014). Moreover, the passenger phenomenon is the simplest explanation for the apparent (but very limited) contribution of thymospheres to TEC subsets upon grafting observed by Ucar et al., without invoking the possibility that ex vivo culture had resulted in changes to cellular function or phenotype. This passenger TEC scenario is supported by our data with purified EpCAM cellderived thymospheres and a lineage tracing strategy, which did not provide evidence for the contribution of EpCAM -derived thymosphere cells to TEC populations. Our lineage tracing and phenotypic analysis establish that TSFCs share a common phenotype and ancestry with thymic fibroblasts that reside in the EpCAM compartment. While we cannot rule out other minor TSFC populations, our data demonstrate that the majority of TSFCs are MTS15+ NCCderived cells and confirm that TSFCs do not belong to the same ontological lineage as the endoderm-derived adult TEC. This finding was corroborated by our microarray expression analysis showing that thymospheres share greater similarity to MSCs than TECs. Although our analyses indicate that TSFCs do not belong to a TEC differentiation hierarchy, they may be important for the maintenance of the thymic stroma. It is well established that mesenchymal fibroblasts provide growth factors (such as fibroblast growth factor 7 [FGF-7], FGF-10, epidermal growth factor [EGF], and retinoic acid) to developing epithelial cells (Jenkinson et al., 2003; Revest et al., 2001; Shinohara and Honjo, 1996; Sitnik et al., 2012) and essential cytokines and extracellular matrix components to thymocyte precursors (Anderson et al., 1997; Gray et al., 2007). Postnatally, fibroblasts, including those in the perivascular region, have been shown to support mTECs in the steady-state thymus and are important for thymic vascularization and regeneration following damage by infection or cyclophosphamide treatment (Lax et al., 2012; Sun et al., 2015). Although the mechanisms of these activities remain obscure,

(C) Maximal intensity projection images of thymospheres generated from purified Lin EpCAM SCA-1+ YFP+ cells from the thymus of 8- to 10-week-old Wnt1cre;R26R-EYFP mice. Data are representative of 3 independent experiments. Scale bar, 25 mm. (D) Representative images of oil red O immunocytochemistry (left) and immunofluorescent anti-FABP4 (green) and DAPI nuclear counterstain (blue) (right) on Lin EpCAM SCA1+ TSFCs cultured for 14 days under maintenance (control) or adipocyte differentiation conditions. Scale bars, 100 mm. Data are representative of 3 independent experiments.

Cell Reports 21, 934–942, October 24, 2017 939

the continued postnatal expression of TEC trophic factors suggests an ongoing role in growth factor provision. It is unknown whether a thymic resident stem or progenitor population maintains the non-epithelial stromal compartment, but our demonstration that TSFCs are mesenchymal with progenitorlike capacities, together with the recent demonstration that TSFCs are long-lived (Ucar et al., 2016), suggests that they are candidates for this role. The regulation of such a population might play a key role in the compositional changes associated with age-related involution (including increased fibroblast and adipocyte presence) and poor thymic function (Taub and Longo, 2005). That this population can be propagated as thymospheres in vitro opens up the possibility that they might be amenable to interrogation in a screening platform to better understand their requirements and role, as has been performed with other sphere-compatible systems (Saxe et al., 2007; Sheridan et al., 2015). EXPERIMENTAL PROCEDURES Unless otherwise stated, all reagents were purchased from Sigma or Thermo Fisher Scientific. Mice Mutant mice were generated on a C57BL/6 or 129/C57BL/6 background and backcrossed at least 10 times to a C57BL/6 background. Foxn1Cre/+ R26REYFP/+, Wnt1Cre/+ R26REYFP/+ and Mesp1Cre/+R26REYFP/+, mice were generated by crossing R26RYFP/YFP (Srinivas et al., 2001) females with Foxn1Cre/+ (Zuklys et al., 2009), Wnt1Cre/+ (Danielian et al., 1998) or Mesp1Cre/+ (Saga et al., 1999) males, respectively. For embryos, noon of the day of vaginal plug detection was taken as embryonic day 0.5 (E0.5). All mice were raised under specific-pathogen-free housing conditions according to the regulations of the Walter and Eliza Hall Institute of Medical Research, the Swiss Federal Government, and the Weatherall Institute of Molecular Medicine. Adult Thymic Stromal Cell Purification and Flow Cytometric Analysis Flow cytometric analysis and/or purification of adult stromal populations were performed as previously described using an optimized protocol with highly purified enzyme preparations and the antibodies detailed in Table S2 (Jain and Gray, 2014; Seach et al., 2012). Placement of regions was guided by fluorescence minus one controls or location of reference populations where appropriate. For a detailed protocol, see Supplemental Experimental Procedures and Figures S1 and S2. Reaggregate Thymic Organ Culture Generation and Grafting Reaggregate organ cultures were generated as described previously (Sheridan et al., 2009). Briefly, thymospheres were mixed with 500,000–1,000,000 BALB/c E15.5 thymic cells with or without 50,000 BALB/c primary embryonic fibroblasts and reaggregated at the gas-liquid interface in RTOC medium (DMEM/F12 with Glutamax supplemented with 10% fetal calf serum [FCS], 15 mM HEPES, 55 mM 2-mercaptoethanol, 13 non-essential amino acids with or without 50 U/mL penicillin, and 50 mg/mL streptomycin). Following overnight incubation, reaggregates were grafted under the renal capsule of 7- to 9-week-old male BALB/c recipient mice and analyzed 4–10 weeks later. Thymosphere Culture Thymosphere cultures were performed as previously described (Ucar et al., 2014). Briefly, cells were plated in ultra-low-attachment plates (Corning) at various densities (1,000–100,000/mL in MEBM medium [Lonza] or DMEM/ F12 supplemented with B27 supplement, 0.5 mg/mL hydrocortisone, and 5 mg/mL insulin [Roche], 4 mg/mL heparin, 20 ng/mL basic FGF [bFGF], and EGF (R&D Systems). Medium was changed every 3 days, and thymospheres were analyzed after 6–8 days.

940 Cell Reports 21, 934–942, October 24, 2017

Immunofluorescence and Immunohistochemistry Thymospheres or cells were fixed in 2% paraformaldehyde (PFA) then permeabilized and blocked in immunofluorescence wash buffer (PBS containing 0.6% Triton X-100) containing 5% (v/v) goat or donkey serum for 30 min. Following incubation with primary antibodies (Table S2) overnight at 4 C, samples were washed and incubated with secondary antibody conjugates (Table S2) for 2 hr before counterstaining with 2 mg/mL DAPI. For imaging, samples were suspended in ProLong Diamond mountant and deposited into coverslipped bottom chamber slides (Ibidi). Adipogenic cultures were fixed with 4% PFA, blocked with donkey serum, and stained with anti-FABP4 primary antibody and anti-goat Alexa 488 antibody and DAPI (Table S2). For immunocytochemistry, adipogenic cultures were fixed with formalin and stained with oil red O as per the manufacturer’s protocols. All images were captured on a LSM780 using Zen 2012 SP2 (black) software v11.0 (Zeiss) or Olympus CKX41 brightfield microscope. Single optical sections and maximal intensity projection images were processed for presentation using OMERO (Allan et al., 2012). Limiting Dilution Analysis Cells were plated in replicate at multiple densities and allowed to grow for 1 week. Wells were imaged on an Eclipse Ti microscope (Nikon) using MetaMorph v7.8.2.0 software (Molecular Devices), and tiled composites were generated using Fiji (scripts by Lachlan Whitehead, Walter and Eliza Hall Institute). Wells were scored for the presence of thymospheres, and the frequency of positive wells at each dilution was used to calculate TSFC frequency using extreme limiting dilution analysis (ELDA) (Hu and Smyth, 2009). Statistical Analysis Statistical analysis was performed using GraphPad Prism software (version 6). Data are reported as mean ± SEM unless otherwise specified. Where applicable, the non-parametric Mann-Whitney U test (two tailed) or unpaired Student’s t test with Welch’s correction was used, and a p value % 0.05 was considered statistically significant. SUPPLEMENTAL INFORMATION Supplemental Information includes Supplemental Experimental Procedures, three figures, and two tables and can be found with this article online at https://doi.org/10.1016/j.celrep.2017.09.090. AUTHOR CONTRIBUTIONS Conceptualization, J.M.S. and D.H.D.G.; Methodology, J.M.S., N.K., S.N.A.R., M.E.R., T.S.P.H., R.J., and J.A.; Investigation, J.M.S., A.K., A.P., N.R., N.K., S.N.A.R., M.E.R., and D.H.D.G.; Writing – Original Draft, J.M.S. and D.H.D.G.; Writing – Review & Editing, J.M.S., A.P., M.E.R., T.S.P.H., J.A., and D.H.D.G. ACKNOWLEDGMENTS We gratefully acknowledge members of the Gray, Strasser, and Herold labs, Dr. S. Taoudi, Prof. J. Visvader, and Prof. A. Strasser for valuable feedback. We thank the WEHI Flow Cytometry Laboratory and Centre for Dynamic Imaging for technical assistance; B. Helbert, K. Mackwell, and C. Young for mouse genotyping; G. Siciliano, K. Humphreys, S. O’Connor, and H. Marks for animal husbandry; Dr. R Boyd for antibodies; and Drs. C. Biben and T. Thomas and Prof. W. Alexander for mice. This work was supported by NHMRC grants GNT0637353 and GNT1049724; NHMRC Career Development Fellowships 1090236 (D.H.D.G.), 1104924 (M.E.R.), and 1107188 (T.S.P.H.); MIRS and MIFRS from the University of Melbourne (R.J.); and the Cancer Council of Victoria (A.P.). This work was made possible through Victorian State Government Operational Infrastructure Support and Australian Government NHMRC IRIISS. Received: January 23, 2017 Revised: June 28, 2017 Accepted: September 26, 2017 Published: October 24, 2017

REFERENCES Allan, C., Burel, J.-M., Moore, J., Blackburn, C., Linkert, M., Loynton, S., Macdonald, D., Moore, W.J., Neves, C., Patterson, A., et al. (2012). OMERO: flexible, model-driven data management for experimental biology. Nat. Methods 9, 245–253. Anderson, G., Anderson, K.L., Tchilian, E.Z., Owen, J.J., and Jenkinson, E.J. (1997). Fibroblast dependency during early thymocyte development maps to the CD25+ CD44+ stage and involves interactions with fibroblast matrix molecules. Eur. J. Immunol. 27, 1200–1206. Bleul, C.C., Corbeaux, T., Reuter, A., Fisch, P., Mo¨nting, J.S., and Boehm, T. (2006). Formation of a functional thymus initiated by a postnatal epithelial progenitor cell. Nature 441, 992–996. Boehm, T., and Swann, J.B. (2013). Thymus involution and regeneration: two sides of the same coin? Nat. Rev. Immunol. 13, 831–838. Boyd, R.L., Tucek, C.L., Godfrey, D.I., Izon, D.J., Wilson, T.J., Davidson, N.J., Bean, A.G., Ladyman, H.M., Ritter, M.A., and Hugo, P. (1993). The thymic microenvironment. Immunol. Today 14, 445–459. Chaudhry, M.S., Velardi, E., Dudakov, J.A., and van den Brink, M.R.M. (2016). Thymus: the next (re)generation. Immunol. Rev. 271, 56–71. Corbeaux, T., Hess, I., Swann, J.B., Kanzler, B., Haas-Assenbaum, A., and Boehm, T. (2010). Thymopoiesis in mice depends on a Foxn1-positive thymic epithelial cell lineage. Proc. Natl. Acad. Sci. USA 107, 16613–16618. Danielian, P.S., Muccino, D., Rowitch, D.H., Michael, S.K., and McMahon, A.P. (1998). Modification of gene activity in mouse embryos in utero by a tamoxifeninducible form of Cre recombinase. Curr. Biol. 8, 1323–1326. Foster, K., Sheridan, J., Veiga-Fernandes, H., Roderick, K., Pachnis, V., Adams, R., Blackburn, C., Kioussis, D., and Coles, M. (2008). Contribution of neural crest-derived cells in the embryonic and adult thymus. J. Immunol. 180, 3183–3189.

Le Douarin, N.M., and Jotereau, F.V. (1975). Tracing of cells of the avian thymus through embryonic life in interspecific chimeras. J. Exp. Med. 142, 17–40. Le Lie`vre, C.S., and Le Douarin, N.M. (1975). Mesenchymal derivatives of the neural crest: analysis of chimaeric quail and chick embryos. J. Embryol. Exp. Morphol. 34, 125–154.  Mayer, C.E., Zuklys, S., Zhanybekova, S., Ohigashi, I., Teh, H.-Y., Sansom, S.N., Shikama-Dorn, N., Hafen, K., Macaulay, I.C., Deadman, M.E., et al. (2016). Dynamic spatio-temporal contribution of single b5t+ cortical epithelial precursors to the thymus medulla. Eur. J. Immunol. 46, 846–856. Moll, R., Divo, M., and Langbein, L. (2008). The human keratins: biology and pathology. Histochem. Cell Biol. 129, 705–733. Nowell, C.S., Bredenkamp, N., Tete´lin, S., Jin, X., Tischner, C., Vaidya, H., Sheridan, J.M., Stenhouse, F.H., Heussen, R., Smith, A.J.H., and Blackburn, C.C. (2011). Foxn1 regulates lineage progression in cortical and medullary thymic epithelial cells but is dispensable for medullary sublineage divergence. PLoS Genet. 7, e1002348. Ohigashi, I., Zuklys, S., Sakata, M., Mayer, C.E., Zhanybekova, S., Murata, S., Tanaka, K., Holla¨nder, G.A., and Takahama, Y. (2013). Aire-expressing thymic medullary epithelial cells originate from b5t-expressing progenitor cells. Proc. Natl. Acad. Sci. USA 110, 9885–9890. Ohigashi, I., Zuklys, S., Sakata, M., Mayer, C.E., Hamazaki, Y., Minato, N., Holla¨nder, G.A., and Takahama, Y. (2015). Adult thymic medullary epithelium is maintained and regenerated by lineage-restricted cells rather than bipotent progenitors. Cell Rep. 13, 1432–1443. Osada, M., Singh, V.J., Wu, K., Sant’Angelo, D.B., and Pezzano, M. (2013). Label retention identifies a multipotent mesenchymal stem cell-like population in the postnatal thymus. PLoS ONE 8, e83024. Pastrana, E., Silva-Vargas, V., and Doetsch, F. (2011). Eyes wide open: a critical review of sphere-formation as an assay for stem cells. Cell Stem Cell 8, 486–498.

Gordon, J., Bennett, A.R., Blackburn, C.C., and Manley, N.R. (2001). Gcm2 and Foxn1 mark early parathyroid- and thymus-specific domains in the developing third pharyngeal pouch. Mech. Dev. 103, 141–143.

€cker, J.C. (2007). Zoned out: functional mapping Petrie, H.T., and Zu´n˜iga-Pflu of stromal signaling microenvironments in the thymus. Annu. Rev. Immunol. 25, 649–679.

Gordon, J., Wilson, V.A., Blair, N.F., Sheridan, J.M., Farley, A., Wilson, L., Manley, N.R., and Blackburn, C.C. (2004). Functional evidence for a single endodermal origin for the thymic epithelium. Nat. Immunol. 5, 546–553.

Revest, J.M., Suniara, R.K., Kerr, K., Owen, J.J., and Dickson, C. (2001). Development of the thymus requires signaling through the fibroblast growth factor receptor R2-IIIb. J. Immunol. 167, 1954–1961.

Gray, D.H.D., Tull, D., Ueno, T., Seach, N., Classon, B.J., Chidgey, A., McConville, M.J., and Boyd, R.L. (2007). A unique thymic fibroblast population revealed by the monoclonal antibody MTS-15. J. Immunol. 178, 4956–4965.

Rodewald, H.R., Paul, S., Haller, C., Bluethmann, H., and Blum, C. (2001). Thymus medulla consisting of epithelial islets each derived from a single progenitor. Nature 414, 763–768.

Gray, D.H.D., Fletcher, A.L., Hammett, M., Seach, N., Ueno, T., Young, L.F., Barbuto, J., Boyd, R.L., and Chidgey, A.P. (2008). Unbiased analysis, enrichment and purification of thymic stromal cells. J. Immunol. Methods 329, 56–66.

Rossi, S.W., Chidgey, A.P., Parnell, S.M., Jenkinson, W.E., Scott, H.S., Boyd, R.L., Jenkinson, E.J., and Anderson, G. (2007). Redefining epithelial progenitor potential in the developing thymus. Eur. J. Immunol. 37, 2411–2418.

Hu, Y., and Smyth, G.K. (2009). ELDA: extreme limiting dilution analysis for comparing depleted and enriched populations in stem cell and other assays. J. Immunol. Methods 347, 70–78. Jain, R., and Gray, D.H.D. (2014). Isolation of thymic epithelial cells and analysis by flow cytometry. Curr. Protoc. Immunol 107, 3.26.1–3.26.15. Jenkinson, W.E., Jenkinson, E.J., and Anderson, G. (2003). Differential requirement for mesenchyme in the proliferation and maturation of thymic epithelial progenitors. J. Exp. Med. 198, 325–332. Jin, X., Nowell, C.S., Ulyanchenko, S., Stenhouse, F.H., and Blackburn, C.C. (2014). Long-term persistence of functional thymic epithelial progenitor cells in vivo under conditions of low FOXN1 expression. PLoS ONE 9, e114842. Komada, Y., Yamane, T., Kadota, D., Isono, K., Takakura, N., Hayashi, S., and Yamazaki, H. (2012). Origins and properties of dental, thymic, and bone marrow mesenchymal cells and their stem cells. PLoS ONE 7, e46436. Lax, S., Ross, E.A., White, A., Marshall, J.L., Jenkinson, W.E., Isacke, C.M., Huso, D.L., Cunningham, A.F., Anderson, G., and Buckley, C.D. (2012). CD248 expression on mesenchymal stromal cells is required for post-natal and infection-dependent thymus remodelling and regeneration. FEBS Open Bio 2, 187–190.

Saga, Y., Miyagawa-Tomita, S., Takagi, A., Kitajima, S., Miyazaki, Ji., and Inoue, T. (1999). MesP1 is expressed in the heart precursor cells and required for the formation of a single heart tube. Development 126, 3437–3447. Saxe, J.P., Wu, H., Kelly, T.K., Phelps, M.E., Sun, Y.E., Kornblum, H.I., and Huang, J. (2007). A phenotypic small-molecule screen identifies an orphan ligand-receptor pair that regulates neural stem cell differentiation. Chem. Biol. 14, 1019–1030. Seach, N., Wong, K., Hammett, M., Boyd, R.L., and Chidgey, A.P. (2012). Purified enzymes improve isolation and characterization of the adult thymic epithelium. J. Immunol. Methods 385, 23–34. Sekai, M., Hamazaki, Y., and Minato, N. (2014). Medullary thymic epithelial stem cells maintain a functional thymus to ensure lifelong central T cell tolerance. Immunity 41, 753–761. Sheridan, J.M., Taoudi, S., Medvinsky, A., and Blackburn, C.C. (2009). A novel method for the generation of reaggregated organotypic cultures that permits juxtaposition of defined cell populations. Genesis 47, 346–351. Sheridan, J.M., Ritchie, M.E., Best, S.A., Jiang, K., Beck, T.J., Vaillant, F., Liu, K., Dickins, R.A., Smyth, G.K., Lindeman, G.J., and Visvader, J.E. (2015). A pooled shRNA screen for regulators of primary mammary stem and progenitor cells identifies roles for Asap1 and Prox1. BMC Cancer 15, 221.

Cell Reports 21, 934–942, October 24, 2017 941

Shinohara, T., and Honjo, T. (1996). Epidermal growth factor can replace thymic mesenchyme in induction of embryonic thymus morphogenesis in vitro. Eur. J. Immunol. 26, 747–752. Sitnik, K.M., Kotarsky, K., White, A.J., Jenkinson, W.E., Anderson, G., and Agace, W.W. (2012). Mesenchymal cells regulate retinoic acid receptordependent cortical thymic epithelial cell homeostasis. J. Immunol. 188, 4801–4809.

tent for medullary and cortical thymic epithelial lineages. Immunity 41, 257–269. Ucar, O., Li, K., Dvornikov, D., Kreutz, C., Timmer, J., Matt, S., Brenner, L., Smedley, C., Travis, M.A., Hofmann, T.G., et al. (2016). A thymic epithelial stem cell pool persists throughout ontogeny and is modulated by TGF-b. Cell Rep. 17, 448–457.

Srinivas, S., Watanabe, T., Lin, C.S., William, C.M., Tanabe, Y., Jessell, T.M., and Costantini, F. (2001). Cre reporter strains produced by targeted insertion of EYFP and ECFP into the ROSA26 locus. BMC Dev. Biol. 1, 4.

Ulyanchenko, S., O’Neill, K.E., Medley, T., Farley, A.M., Vaidya, H.J., Cook, A.M., Blair, N.F., and Blackburn, C.C. (2016). Identification of a bipotent epithelial progenitor population in the adult thymus. Cell Rep. 14, 2819– 2832.

Sun, L., Sun, C., Liang, Z., Li, H., Chen, L., Luo, H., Zhang, H., Ding, P., Sun, X., Qin, Z., and Zhao, Y. (2015). FSP1(+) fibroblast subpopulation is essential for the maintenance and regeneration of medullary thymic epithelial cells. Sci. Rep. 5, 14871.

Wong, K., Lister, N.L., Barsanti, M., Lim, J.M.C., Hammett, M.V., Khong, D.M., Siatskas, C., Gray, D.H.D., Boyd, R.L., and Chidgey, A.P. (2014). Multilineage potential and self-renewal define an epithelial progenitor cell population in the adult thymus. Cell Rep. 8, 1198–1209.

Taub, D.D., and Longo, D.L. (2005). Insights into thymic aging and regeneration. Immunol. Rev. 205, 72–93.

Zuklys, S., Gill, J., Keller, M.P., Hauri-Hohl, M., Zhanybekova, S., Balciunaite, G., Na, K.-J., Jeker, L.T., Hafen, K., Tsukamoto, N., et al. (2009). Stabilized beta-catenin in thymic epithelial cells blocks thymus development and function. J. Immunol. 182, 2997–3007.

Ucar, A., Ucar, O., Klug, P., Matt, S., Brunk, F., Hofmann, T.G., and Kyewski, B. (2014). Adult thymus contains FoxN1(-) epithelial stem cells that are bipo-

942 Cell Reports 21, 934–942, October 24, 2017