FSP1+ fibroblast subpopulation is essential for the maintenance and regeneration of medullary thymic epithelial cells

Thymic epithelial cells (TECs) form a 3-dimentional network supporting thymocyte development and maturation. Besides epithelium and thymocytes, heterogeneous fibroblasts are essential components in maintaining thymic microenvironments. However, thymic fibroblast characteristics, development and function remain to be determined. We herein found that thymic non-hematopoietic CD45-FSP1+ cells represent a unique Fibroblast specific protein 1 (FSP1)—fibroblast-derived cell subset. Deletion of these cells in FSP1-TK transgenic mice caused thymus atrophy due to the loss of TECs, especially mature medullary TECs (MHCIIhigh, CD80+ and Aire+). In a cyclophosphamide-induced thymus injury and regeneration model, lack of non-hematopoietic CD45-FSP1+ fibroblast subpopulation significantly delayed thymus regeneration. In fact, thymic FSP1+ fibroblasts released more IL-6, FGF7 and FSP1 in the culture medium than their FSP1- counterparts. Further experiments showed that the FSP1 protein could directly enhance the proliferation and maturation of TECs in the in vitro culture systems. FSP1 knockout mice had significantly smaller thymus size and less TECs than their control. Collectively, our studies reveal that thymic CD45-FSP1+ cells are a subpopulation of fibroblasts, which is crucial for the maintenance and regeneration of TECs especially medullary TECs through providing IL-6, FGF7 and FSP1.

factor; KGF) and FGF10 17,19,20 . Thus, the development and maturation of TECs critically depend on the complicate microenvironments, mainly offered by residual surrounding cells such as immune cells and fibroblasts.
Fibroblast heterogeneity has been appreciated for several decades [21][22][23] , but its biological significance and the basis for cellular diversity remain uncertain. At present, ER-TR7 and MTS-15 are considered as specific markers for thymic fibroblasts 16,24 . However, markers for thymic fibroblasts are easily confusing with mesenchymal cells 25 . Fibroblast-specific protein 1 (FSP1, also named as S100A4), one member of the S100 superfamily of cytoplasmic calcium-binding proteins, is predominately expressed in fibroblasts but not in epithelial cells in organs undergoing tissue remodeling like skin, kidney, lung, and heart, as well as in some other cell types in certain conditions [26][27][28][29] . The presence, characteristics and biological significance of non-hematopoietic FSP1 + cells in the thymus have not been determined. In the present study, using FSP1-GFP reporter mice, FSP1 + cells-deleting mice (FSP1-thymidine kinase (TK) transgenic mice), FSP1 knockout (FSP1KO) mice, and many experimental mouse models, we tried to investigate the characteristics and biological significance of non-hematopoietic FSP1 + cells in the thymus. We found that a subpopulation of fibroblasts but no epithelial cells express FSP1 in the thymus. A series of in vivo and in vitro studies indicated that non-hematopoietic CD45 − FSP1 + fibroblast subpopulation plays an important nursing role on TEC maintenance and regeneration via providing IL-6, FGF7 and FSP1. The present study shed lights on the critical roles of FSP1 + fibroblast subset and FSP1 on mTEC development.

Results
Thymic CD45 -FSP1 + cells are a subpopulation of fibroblasts. FSP1 was originally recognized as a specific marker for fibroblasts 26 . However, it was recently challenged by the observation showing the expression of FSP1 in other cells in inflammatory situations 30 . Considering the fibroblast heterogeneity and the differences of fibroblasts in different organs 16,[21][22][23] , we firstly investigated the expression pattern of FSP1 in different cell types in the thymus using immunohistochemical staining assays. Immunofluorescence analysis of adult mouse thymus sections with anti-FSP1 antibody revealed specific and extensive staining (Fig. 1A). The staining patterns of FSP1 in thymic medulla and cortex regions were different. FSP1 was expressed intensively and distributed clusteredly in medulla area, whereas FSP1 in cortex area was less and point shape distribution (Fig. 1A). Co-staining of FSP1 and mTEC marker UEA-1 or MHCII showed majority of FSP1 + cells were located in thymic medullary area (Fig. 1B). Because CD31, known as platelet/endothelial cell adhesion molecule-1, is widely recognized and frequently used as a sensitive and relatively specific immunohistochemical marker of endothelial cells and thereby vascular neoplasia 31 , we thus investigated whether CD31 + cells express FSP1 in the thymus. As shown in Fig. 1C, no CD31 + cells were co-stained with FSP1. In addition, no FSP1 + cells in the thymus express α -smooth muscle actin (α -SMA) (Fig. 1C), one of the mesenchymal markers 32 . To further determine the expression pattern of FSP1 in the thymus, we used FSP1-GFP reporter mice 33 , in which the FSP1 promoter drives green fluorescent protein (GFP) expression, to mark and trace FSP1 + cells. No CD45 − FSP1 + cells in the thymus express CD31 in WT and FSP-GFP reporter mice (Fig. 1C,D). Clearly, TECs including CD45 − EpCAM + TECs, CD45 − EpCAM + UAE-1 + mTECs, and CD45 − EpCAM + BP-1 + cTECs of FSP-GFP reporter mice did not express detectable FSP1 as determined by flow cytometry (Fig. 1D,E). The undetectable expression of FSP1 in TECs was also confirmed by real-time PCR with FSP1 + and FSP1 − fibroblasts as positive and negative control, respectively (Fig. 1F). The poor FSP1 expression in TECs was in line well with a recent report 34 . However, part of CD45 − FSP1 + cells isolated from FSP1-GFP reporter mice showed MTS15 + , one marker for thymic fibroblasts 16 , as determined by flow cytometry (Fig. 1D). In addition, CD45 − FSP1 + cells expressed thymic mesenchymal cell markers BP-1 and gp38, and the majority of CD45 − FSP1 + cells expressed PDGFRα and PDGFRβ (Fig. 1G,H), indicating their mesenchymal origin. Furthermore, immunohistological staining of the thymus showed that only a small fraction of FSP1 + cells were co-localized with MTS15 and ER-TR7, respectively (Fig. 1I). More impressively, the distribution of MTS15 + and ER-TR7 + fibroblasts in the thymus was different from the distribution of FSP1 + cells (Fig. 1I). To further confirm whether thymic CD45 − FSP1 + cells were fibroblasts, we cultured mouse primary thymic fibroblasts in vitro. They displayed classical spindle shape and expressed FSP1 (Suppl. Fig. 1A) and fibroblast markers vimentin, MTS15 and PDGFRα (Fig. 1J, Suppl. Fig. 1B). Furthermore, the cultured thymic FSP1 + cells did not express pan CK, CD11b, UEA-1, CD11c, and F4/80, markers for other cell types (Suppl. Fig. 1C). All these results provided evidence that thymic CD45 − FSP1 + cells were not TECs, endothelial cells or perivascular smooth muscle cells, but likely were a unique subgroup of fibroblasts, which was different from the known MTS15 + and ER-TR7 + fibroblasts 16 .
In order to investigate whether FSP1 − thymic fibroblasts could transfer into FSP1 + cells or vise verse, we cultured the sorted FSP1-GFP − and FSP1-GFP + thymic fibroblasts of FSP1-GFP reporter mice in vitro. We surprisingly found that the sorted FSP1-GFPfibroblasts could generally change into FSP1-GFP + cells in a time-dependent manner (Fig. 1K). Although FSP1-GFP − fibroblasts could turn into FSP1-GFP + cells, while FSP1-GFP + cells stayed FSP1 positive status (Supplemental Fig. 1D,E). Moreover, this change of FSP1 expression is independent on cell proliferation. Cell proliferation was inhibited when the isolated thymic fibroblasts were treated with mitomycin C (MitoC) for 4 hrs and were then cultured for following 4 days (Suppl. Fig. 1F). When thymic fibroblasts of FSP1-GFP mice were treated with MitoC for 4 hrs after 4 days culture, these MitoC-treated FSP1-GFP − thymic fibroblasts could also Scientific RepoRts | 5:14871 | DOi: 10.1038/srep14871 turn into FSP1-GFP + cells efficiently as cells without MitoC treatment (Fig. 1L). In addition, thymic FSP1-GFP − and FSP1-GFP + fibroblasts were different in their cell size. Thymic FSP1-GFP − fibroblasts were smaller than FSP1-GFP + fibroblasts as determined by their forward scatter signal when cells were assayed by flow cytometry (data not shown). Thus, thymic FSP1 + fibroblasts were a fraction of thymic fibroblasts which could be derived from FSP1 − fibroblasts.
Deletion of thymic FSP1 + fibroblasts impairs thymic maintenance. To investigate the role of FSP1 + cells on the maintenance of TECs, FSP1-TK transgenic mice were used 35,36 , in which the expression of herpes simplex virus-derived thymidine kinase (TK) was under the control of the FSP1 promoter. Exogenous addition of oligonucleotide analogue, such as ganciclovir (GCV), phosphorylated product by TK could selectively delete proliferating FSP1 + cells in vivo 35 . In our current study, FSP1-TK mice (TK + ) and wild-type (WT) control littermates (TK − ) were treated with GCV for 18 days, thymic FSP1 expression was remarkably decreased in TK + mice as determined by immunofluorescence (Fig. 2C). Compare to the GCV-treated TK − control mice, GCV-treated TK + mice had small thymus size, decreased thymus weight and total cell number of thymocyte (P < 0.001, Fig. 2A,B, and suppl. Fig. 2A). Thymocytes were severely affected with almost complete loss of thymocyte subsets (suppl. Fig. 2B). Thymic sections from TK − and TK + mice stained with H&E or mAbs against CK5 and CK8 revealed dramatic disruption of thymic structure with confusing thymic medulla and cortex region (Fig. 2C). The total cell number of thymic epithelial cells (CD45 − EpCAM + ) were also significantly decreased (Fig. 2D). Moreover, the (L) Percentage of FSP1 + cells in primary thymic fibroblasts with or without 10 μ g/ml mitomycin C treatment for 4 hours. Representative results are shown from one of three independent experiments performed. percentage of TECs expressing high level of MHCII were decreased while MHCII low TECs were relatively increased, although their cell number were significantly decreased (P < 0.001, Fig. 2E). We further analyzed TEC subsets by using UEA-1 and BP-1 as surface markers of mTEC and cTEC subpopulation respectively 6 . The results showed that both the ratio and cell number of mTECs (CD45 − EpCAM + UEA-1 + ) were remarkably decreased in TK + mice (P < 0.001, Fig. 2F,G). However, the total cell number of cTECs (CD45 − EpCAM + BP-1 + ) was indistinguishable in TK + mice and TK − mice (Fig. 2G). The increased percentage of cTECs in TK + mice was possibly caused by the relative decrease of mTEC components (Fig. 2F,G). Thus, deletion of thymic FSP1 + cells selectively affected mTECs in the thymus. Functional maturation of mTECs is marked by expressing high level of MHC II, CD80 and the transcriptional regulator Aire 6,37 . The mature mTECs (MHCII high , CD80 high and Aire + ) were significantly decreased in GCV-treated TK + mice compared with GCV-treated TK − mice ( Fig. 2H-J). These results suggested that deletion of FSP1 + cells selectively impaired the mature mTEC homeostasis.
Others and our studies (data not shown) showed that some subsets of immune cells like macrophages also expressed FSP1 in certain situations 30,38 , so it is essential to identify which cell type of thymic FSP1 + cells offered the supporting effects on mTEC homeostasis. We employed full bone marrow chimeric (A) Representative photographs of thymus organs in TK − and TK + transgenic mice with and without GCV treatment for 18 days were shown. (B) Thymus weight and total cell numbers of thymocytes in TK − and TK + mice with GCV treatment were summarized. (C) Thymic sections from GCV treated TK − and TK + mice were stained with FSP1 to determine the ablating efficiency (upper). The H&E and CK5/CK8 staining of thymic sections showed decreased area of thymic medullary region in TK + mice than in TK − mice after GCV treatment (middle and lower). (D) Representative FACS analysis of CD45 − EpCAM + TECs, MHCII high and MHCII low TECs in the isolated thymic cells was shown. The cell number of TECs in TK + mice was less than in TK − mice after GCV treatment. (E) The percentage and the total cell number of MHCII high and MHCII low TECs in GCV-treated TK − and TK + mice were summarized. Representative FACS profiles (F), and the percentage and total cell number (G) of mTECs and cTECs in CD45 − EpCAM + cells in GCV-treated TK − and TK + mice were shown. Phenotypic characterization (H) and the percentages (I) of MHCII + , CD80 + and Aire + mTECs in the gated thymic CD45 − or CD45 − EpCAM + cells of TK − and TK + mice with GCV treatment. (J) The total cell numbers of CD45 − EpCAM + UEA-1 + MHCII high , CD45 − EpCAM + UEA-1 + CD80 high , CD45 − EpCAM + UEA-1 + CD80 low and CD45 − EpCAM + UEA-1 + Aire + cells in GCV-treated TK − and TK + mice. Representative results are shown from one of three independent experiments performed. Data were shown as mean ± SD (N = 5). *P < 0.05, **P < 0.01, ***P < 0.001 compared with TK − mice.
Scientific RepoRts | 5:14871 | DOi: 10.1038/srep14871 mouse models to elucidate whether FSP1 + immune cells participated in the impaired mTEC maintenance in GCV-treated FSP1-TK mice. First, we adoptively transferred either TK − or TK + bone marrow cells (BMCs) into lethally irradiated TK − mice to establish the full chimeras ( Fig. 3A) 39 . After 8 weeks after transplantation, recipient mice were treated with GCV for 18 days and the thymocyte and TEC subsets were assayed. In TK − mice received TK + BMCs, CD45 + FSP1 + cells were from donors and would be deleted by GCV treatment, while thymic CD45 − FSP1 + cells, which were from recipient mice, would not be deleted by GCV treatment. The identical thymus size, the ratio of thymus weight to body weight, total cell number, percentages of thymocyte, TECs, mTECs, cTECs and mature mTECs were observed in recipients receiving TK − and TK + BMCs respectively ( Fig. 3B-D, and data not shown), indicating that ablation of hematopoietic-derived FSP1 + cells, did not impact TECs in this model. Reversely, another group of full chimeric mice were generated by transplanting TK − BMCs into lethally irradiated either TK − or TK + mice (Fig. 3E). In this model, only non-hematopoietic FSP1 + cells in TK + recipients received TK − BMCs could be deleted after GCV treatment. Notably, smaller thymus size and lower ratio of thymus weight to body weight (P < 0.01, Fig. 3F), and altered total cell number and thymocytes in TK + recipients of TK − BMCs after GCV treatment compared with TK − recipients of TK − BMCs (data not shown). Furthermore, the percentages of mTECs and MHCII high mature TECs but not cTECs were significantly decreased in GCV-treated TK + recipients (P < 0.05, Fig. 3G). Moreover, the percentages were generated by transplanting TK − BMCs to lethally irradiated either TK − and TK + mice. By 8 weeks after transplantation of BMCs, recipient mice were treated with GCV for 18 days, and the TEC subsets were assayed. (F) Representative photographs of thymus organs and the ratio of thymus weight to body weight in TK − and TK + recipient mice received TK − BMCs. (G) Representative FACS staining and the frequencies of TECs, MHCII high and MHCII low TECs, mTECs and cTECs in TK − and TK + recipient mice. (H) Representative FACS staining and the frequencies of CD80 high , CD80 low and Aire expression in mTECs of TK − and TK + recipient mice were shown. Data presented are mean ± SD (N = 5). Representative results are shown from one of three independent experiments performed. *P < 0.05, **P < 0.01, ***P < 0.001 compared with TK − mice. and cell number of mature EpCAM + UEA-1 + mTECs including CD80 high , CD80 + and Aire + cells were markedly decreased (P < 0.05, Fig. 3H and data not shown). Based on all these data, we concluded that non-hematopoietic FSP1 + cells, which were proven to be one of thymic fibroblast subpopulations, played an important role in maintaining TEC particularly mTEC homeostasis in mice.
Thymic FSP1 + fibroblasts promote mTEC regeneration. To assess the roles of thymic FSP1 + fibroblasts in thymic regeneration, we applied thymus regeneration model to detect the recovery efficiency of TEC subsets in the FSP1 + cell-deleted mouse model. Immunosuppressive agent cyclophosphamide (Cy) was used to induce thymus atrophy in TK − and TK + mice as reported previously 40 . GCV was injected for 14 days to delete proliferating FSP1 + cells during thymus recovery, thymic cellular composition and phenotype were then detected at indicated time points (Fig. 4A). In WT mice, Cy treatment induced severe thymus damage within 3 days, then recovery commenced, reaching almost normal levels by day 14 (Fig. 4B and suppl. Fig. 3A). However, deletion of FSP1 + cells by GCV in TK + mice (Fig. 4C, the upper panel) caused significantly delayed thymus regeneration, as indicated by the observation showing that their thymus weight was particularly lower than those in TK − mice at 14 and 21 days, although they could eventually recover to normal level at 28 days ( Fig. 4B and suppl. Fig. 3A). Immunofluorescence analysis of thymus section from 14 days after Cy and GCV treatment stained with CK5 and CK8 revealed a significant decrease in medullary area in TK + mice than TK − mice (Fig. 4C, the lower panel). Meanwhile, the recovery ratio of the thymus weight to body weight, total thymocyte number and cell number of thymocyte subsets in TK + mice were remarkably delayed (suppl. Fig.  3B-D). Importantly, the CD45 − EpCAM + TECs including MHCII high and MHCII low subsets in TK + mice The recovery of mTECs of MHCII high , Aire + , CD80 + and MHCII low in Cy/GCV-treated TK − and TK + mice at indicated time points were present. Data presented are the mean ± SD (N = 6), representing one representative of three independent experiments with identical results. *P < 0.05, **P < 0.01 and ***P < 0.001 compared with TK − controls. recovered in a slower manner than control mice (Fig. 4D). mTECs including relatively mature ones like MHCII high , CD80 + and Aire + cells were unable to be recovered efficiently in TK + mice as control mice after Cy and GCV treatment (Fig. 4D,E). Among them, Aire + mTECs in TK + mice could not recovered to normal level even 28 days after treatment (Fig. 4E). In contrast to mTECs, the total cell number of cTEC including MHCII high and MHCII low cells in TK + mice were similar as in TK − mice (suppl. Fig.  3E-G). Thus, these results demonstrated that FSP1 + cells were essential for mTEC regenerative potentiality after thymus injury.
Furthermore, we applied full bone marrow chimera to address whether hematopoietic or non-hematopoietic FSP1 + cells play the critical role in the thymus regeneration. 8 weeks after transplantation of either TK − or TK + BMCs into lethally irradiated TK − mice, recipient mice were treated with Cy and GCV as same as thymus regeneration model (suppl. Fig. 4A) and analyzed 14 days after Cy treatment. Identical ratio of the thymus weight to body weight, total cell number, the percentage of thymocytes, TECs, mTECs and mature mTECs were observed in recipients receiving both TK − and TK + BMCs (suppl. Fig. 4B-F), suggesting that hematopoietic-derived FSP1 + cells did not impact thymic regenerative capacity. However, in full chimeric mice generated by transplanting TK − BMCs to either TK − or TK + mice (suppl. Fig. 4G), we discovered significantly decreased ratio of thymus weight to body weight and total cell number in TK + recipients of TK − BMCs in which only CD45 − FSP1 + cells were deleted (suppl. Fig. 4H). Notably, the percentage of CD4 + CD8 + thymocytes, and MHCII high , CD80 + and Aire + mature mTECs were also significantly decreased in TK + recipients (suppl. Fig. 4I-L). These results indicated that thymic FSP1 + fibroblasts are important for regulating mTEC regenerative potentiality.
Thymic FSP1 + cells control TEC proliferation. Cy-induced thymic atrophy involved dramatic apoptosis of thymocytes and TECs, thus thymic regeneration depends on constant cell proliferation. Within thymus regeneration model, we postulated that defects in cell proliferation might contribute to the delayed recovery of TECs and mTECs in FSP1 + cell deleted mice. To address this possibility, we detected the proliferative marker Ki67 expression in TECs and mTECs at different time point after Cy and GCV treatment to investigate the cell proliferation rate of TECs and mTECs. According to the dynamics of thymus regeneration, we detected their proliferative capacity at 7, 14 and 21 days after Cy treatment. The percentage of Ki67 + cells was significantly lower in TECs and UEA-1 + mTECs of TK + mice compared with the controls (Fig. 5A-D). Moreover, the absolute cell number of Ki67 + TECs and mTECs in TK + mice were significantly lower than those in TK − mice after Cy and GCV treatment during recovery (Fig. 5E,F). However, the percentages and cell numbers of Ki67 + cTECs were similar in TK − and TK + mice after Cy and GCV treatment during recovery (Fig. 5G-I). Thus, deletion of FSP1 + fibroblasts impaired the cell proliferation of mTECs after injury, which might contribute to the defects of thymic regeneration in FSP1 + cells-deleted mice.
Thymic FSP1 + fibroblasts promote TEC proliferation via IL-6 and FGF7. In order to address the potential mechanisms involved in the regulatory roles of thymic FSP1 + fibroblasts in maintaining thymic microenvironment and promoting thymus regeneration, we used the in vitro cultured thymic fibroblasts in re-aggregated thymic organ culture (RTOC) system. In RTOC, murine TECs and thymocytes were aggregated with or without mitomycin C-treated thymic FSP1 + fibroblasts. After 5 days culture, the results showed that thymic FSP1 + fibroblasts could remarkably enhance MHCII, CD80 and Aire expression on TECs (Fig. 6A), suggesting that thymic FSP1 + fibroblasts could promote TEC maturation.
Thymic fibroblasts are usually considered as nutritious cells to promote thymocyte and TEC proliferation by providing immune modulators. To identify the molecular mechanisms that FSP1 + fibroblasts regulate TEC maintenance and regeneration, we examined the expression profile of a series of cytokines which are particularly critical for TEC proliferation in thymic FSP1 − and FSP1 + fibroblasts 16,41,42 . Real-time PCR results revealed that, among the detected molecules, strikingly high signal for IL-6 and FGF7 were obtained in FSP1 + fibroblasts compared to FSP1 − group (P < 0.001, Fig. 6B). These results were further confirmed by ELISA assays detecting cell culture supernatants from FSP1 − and FSP1 + fibroblasts (P < 0.001, Fig. 6C). IL-6 and FGF7 have been demonstrated to be very important growth factors for TEC proliferation in vitro and in vivo 16,43,44 . When we performed fetal thymus organ culture (FTOC), addition of IL-6 and FGF7 significantly increased total cell number and TEC cell number in the cultured thymic lobes respectively (Fig. 6D,E), confirming that IL-6 and FGF7 markedly supported TEC proliferation 44 .
Apart from locating in the nucleus and cytoplasm, FSP1 was recently found to be released to extracellular space 45,46 . By interacting with their cell surface receptors, FSP1 possesses a wide range of biological functions, such as regulation of cell survival and proliferation 45 . To determine whether FSP1 could be secreted by FSP1 + fibroblasts, we also detected FSP1 protein levels in cell culture supernatants by ELISA. Clearly, thymic FSP1 + fibroblasts could secrete FSP1 protein into the culture medium (P < 0.001, Fig. 6C), leading us to examine the contribution of FSP1 protein in TEC proliferation and function.
Thymic FSP1 + fibroblasts enhance TEC maturation via FSP1. To better comprehend the effect of FSP1 protein in TEC development and function. We investigated the thymus and TEC phenotype of FSP1KO mice. These FSP1KO mice were fertile and displayed no gross abnormalities within 3 months after birth. Immunofluorescence analysis of thymic sections from WT and FSP1KO mice revealed that Scientific RepoRts | 5:14871 | DOi: 10.1038/srep14871 FSP1 expression was eliminated clearly (Fig. 7A). Moreover, FSP1KO mice had a thymus defect as evidenced by the smaller thymus size, lower thymus weight and lower ratio of thymus weight to body weight (P < 0.001, Fig. 7B,C) compared with age-matched WT mice. The total cell number of thymocytes and TECs were significantly decreased in FSP1KO mice (P < 0.001, Fig. 7D). The percentage and cell number of mTECs but not cTECs were remarkably decreased in FSP1KO mice than those in WT mice (P < 0.05, Fig. 7E). Consistently, FSP1KO mice had significantly smaller medulla compared with WT mice, as shown in the thymic sections stained with CK5 (Fig. 7F). The cell number of mTECs including MHCII high , CD40 + and CD80 + cells were markedly decreased in FSP1KO mice, though their percentage were unchanged (Fig. 7G,H). These observations suggested that FSP1 played an important role in maintaining mTEC compartment and enhancing thymic size and cellularity.
More impressively, we found that FSP1 was involved in thymic epithelium regeneration after thymus damage. We detected the recovery efficiency of TEC subsets in FSP1KO and WT mice treated with Cy. Because FSP1KO mice had smaller thymus and lower cellularity than WT mice before Cy treatment, we thus utilized the recovery ratio of cell number to determine the thymic regeneration ability in these assays. By 14 days after Cy treatment, the recovery ratio of the thymus weight, total cell number and thymocyte subsets in FSP1KO mice were strikingly lower than WT mice (P < 0.001, Fig. 7I,J). Notably, mTECs, but not cTECs, and mature mTECs including MHCII high , CD40 + and CD80 + cells in FSP1KO mice were also unable to be recovered efficiently as WT mice (P < 0.001, Fig. 7K). These data indicated that FSP1 deficiency significantly affected the regenerative potentiality of mature mTECs after thymus damage. It has been reported that FSP1 could interact with cell surface receptors such as receptor for advanced glycation end products (RAGE), annexin II, and heparan sulfate proteoglycans 45 . To address the possibility that FSP1 directly regulates TEC differentiation, we first investigated the expression of cell surface receptors for FSP1 in TECs. Real-time PCR assays of the cultured TECs revealed that several FSP1 receptors were expressed in TECs, among which annexin II and syndecan 1, one of heparan sulfate proteoglycans, expressed relatively high (Fig. 7L). We next assessed the potential role of FSP1 protein involved in regulating TECs' differentiation. Importantly, addition of purified FSP1 protein significantly increased the total cell number of TECs, the percentage of Ki67 + cells in TECs and mTECs, as well as the expression of the mature markers CD80 and Aire on mTECs in FTOC culture system (Fig. 7M,N), indicating the supporting role of FSP1 on mTEC proliferation and maturation. Consistently, in the in vitro TEC culture system, FSP1 protein could significantly increase MHCII and CD40 expression on TECs (Fig. 7O). Meanwhile, FSP1 treatment significantly increased the key TEC regulator Foxn1 expression in TECs as detected by Real-time PCR assay (Fig. 7P), which might contribute to the enhancement of MHCII and CD40 expression on TECs. Additionally, FSP1 also increased the expression of FGFR2IIIb, the receptor for FGFs 17 in TECs (Fig. 7P), indicating that FSP1 could promote the ability of TECs utilizing FGFs. Therefore, these data illustrated that thymic FSP1 + fibroblasts could enhance TEC proliferation and differentiation via directly producing FSP1 protein.

Discussion
As the primary lymphoid organ, thymus is the place for T cell development and maturation, playing a crucial role in establishment of T cell immunity and self-tolerance. Thymus possesses a complex microenvironments consisting of many cell components in which TECs and thymocytes are the most abundant and important compositions. Besides, fibroblasts are also an irreplaceable stromal cell type in thymic microenvironments by supporting thymic structure and function 1 . However, due to the heterogeneity The total cell numbers of thymic lobes cultured with IL-6 (100 ng/ml) and FGF7 (100 ng/ml) for 6 days in FTOC were summarized. (E) The total cell numbers of CD45 − EpCAM + TECs in thymic lobes cultured with IL-6 (100 ng/ml) and FGF7 (100 ng/ml) for 6 days in FTOC were summarized. Representative results are shown from one of three independent experiments performed. Data presented are mean ± SD (N = 6). *P < 0.05, **P < 0.01 and ***P < 0.001 (FSP1 − vs FSP1 + or between the indicated groups).
Scientific RepoRts | 5:14871 | DOi: 10.1038/srep14871 and lack of specific markers 25,47,48 , the identification, development and function of thymic fibroblasts remain poorly understood. In our present study, we revealed that non-hematopoietic CD45 − FSP1 + cells in the thymus represent a subpopulation of fibroblasts as supported by the following evidences and characteristics: 1) The cultured primary thymic CD45 − FSP1 + cells displayed classical spindle shape and express fibroblast marker vimentin 49 , but not other cell type markers like cytokeratin, CD11b, CD11c and UEA-1. 2) Thymic CD45 − FSP1 − fibroblasts were smaller but CD45 − FSP1 + cells were larger. 3) In the primary cell culture, thymic CD45 − FSP1 − fibroblasts underwent transformation to CD45 − FSP1 + cells, but no CD45 − FSP1 + cells turn into CD45 − FSP1 − cells. This observation also indicates that thymic CD45 − FSP1 + fibroblasts likely represent a more mature state than CD45 − FSP1 − fibroblasts. 4) Thymic CD45 − FSP1 + fibroblasts produced significantly more IL-6 and FGF7 than CD45 − FSP1 − cells. 5) Little co-location of FSP1 with cytokeratin, CD31 and α -SMA in the thymus assayed by immunohistochemical staining and FSP1-GFP reporter assays implied that thymic CD45 − FSP1 + cells are not epithelium, The recovery ratio of cTECs and mTECs, and mTECs expressing MHCII high , CD40 + and CD80 + in WT and FSP1KO mice were summarized. Data is mean ± SD (4-6 mice/group) from one of two independent experiments. (L) The mRNA levels of RAGE, annexin II and heparan sulfate proteoglycans in cultured primary TECs and thymocytes were assayed by Real-time PCR. (M) FSP1 significantly increased the total cell number of TECs and the percentage of Ki67 + cells in TECs and mTECs after the fetal thymus was cultured with FSP1 in FTOC system for 6 days were summarized. (N) Representative FACS and the percentage of CD80 and Aire expression on mTECs were significantly increased after thymi were cultured in FTOC system with FSP1 for 6 days. (O) Representative FACS and the percentage of MHCII and CD40 expression on TECs were significantly increased after primary TECs were cultured with FSP1 for 5 days. (P) The expression of Foxn1, Wnt4, BMP4 and FGFR2IIIb in TECs cultured with or without FSP1 were detected by Real-time PCR. Data is mean ± SD (3-5 sample/group) from one of two independent experiments. *P < 0.05, **P < 0.01, and ***P < 0.001 (FSP1KO vs WT or between the indicated groups).
Scientific RepoRts | 5:14871 | DOi: 10.1038/srep14871 endothelium or perivascular smooth muscle cells. 6) Immunofluorescence staining of the thymus showed that only a fraction of FSP1 + cells were co-localized with MTS15 and ER-TR7, respectively. Thymic FSP1 + fibroblasts extensively exist in the thymic medullary area, which is different from the distribution MTS15 + and ER-TR7 + fibroblasts. In addition, the majority of CD45 − FSP + fibroblasts express mesenchymal cell markers BP-1, gp38, PDGFRα and PDGFRβ 47,50,51 , indicating that thymic CD45 − FSP1 + fibroblasts were derived from mesenchymal precursors. The previous view of the origin for thymic fibroblast holds that they are developed from neural crest (NC)-derived mesenchyme 52,53 . In a recent study, Komada et al. using double-transgenic mice defined that thymic mesenchymal PDGFRα and PDGFRβ expressing cells were composed of both NC and mesoderm-derived cells, contributing to perivascular cells and endothelial cells, respectively 54 . However, our study showed that thymic CD45 − FSP + fibroblasts express PDGFRα and PDGFRβ but not α -SMA and CD31. Thus, the precise origins of these FSP1 + fibroblasts are still need future investigation 52,53 . All these findings shed lights on the heterogeneity of thymic fibroblasts.
Thymic fibroblasts were considered to play a role in supporting thymus structure by distribution at thymic subcapsule, septae, and near vasculature, and also promoting thymic cellularity by producing secretory mediators 16 . However, few studies directly investigated the function of fibroblasts themselves in the thymus, but indirectly studied their mesenchymal precursors 19,42 or cytokines executors such as FGFs 44,53,55 . Our research using FSP1 + cells-deleted mice (FSP1-TK), full bone marrow chimera mice, combined with thymus regeneration models provide evidence on the function of thymic FSP1 + fibroblasts. We demonstrated that CD45 − FSP1 + cells played an essential role in thymus maintenance as verified by small thymus size, decreased thymus weight and total cell number after systemic ablation of FSP1 + cells. Importantly, FSP1 + cells were required for mTEC homeostasis because of decreased mature mTECs including MHCII high , CD80 + , Aire + cells in FSP1 + cells-deleted thymus. Moreover, deletion of FSP1 + cells significantly impaired thymus full recovery after Cy-induced injury. During the course of thymus regeneration, we found the recovery of thymus weight, thymocytes and TECs, particularly mature mTECs, was remarkably delayed in FSP1 + cells-deleted mice. To avoid the effect of hematopoietic FSP1 + cells, we performed studies using full bone marrow chimeras and concluded that non-hematopoietic CD45 − FSP1 + cells which were proven to be fibroblasts population played the critical role in the thymus maintenance and regeneration. We failed to observe the detectable impacts of hematopoietic FSP1 + cells on TEC maintenance and regeneration in physiological and pathological situations.
Deletion of thymic FSP1 + fibroblasts significantly impacted mTEC proliferation as supported by the decreased Ki67 expression during thymic regeneration. Since many cytokine mediators, which were produced mostly by thymic fibroblasts, were involved in TEC proliferation 16,43,44 , Real-time PCR and ELISA results verified that thymic FSP1 + fibroblasts produced high level of IL-6 and FGF7 than thymic FSP1 − fibroblasts. IL-6 and FGF7 have been demonstrated to be very important growth factors for TEC proliferation in vitro and in vivo 16,43,44 . Our data showed that exposure to IL-6 and FGF7 significantly increased total cell number and TEC cell number of the cultured thymic lobes in FTOC culture, confirming that IL-6 and FGF7 markedly supported TEC proliferation 55 . Therefore, thymic FSP1 + fibroblasts control TEC maintenance and regeneration through their ability to produce large amount of IL-6 and FGF7. FSP1, known as S100A4, belonging to S100 family has been extensively investigated in tumorigenesis. A wealth of information illustrated that FSP1 promoted cancer progression by enhancing cell proliferation, motility, invasiveness, metastasis and angiogenesis 45,[56][57][58] . Except for location in nucleus and cytoplasm, FSP1 was revealed to be secreted into extracellular space to exert their effects by interacting with the cell surface receptors 45,46,59 . Our results showed that primary thymic FSP1 + fibroblasts could release FSP1 into the culture medium, offering the possibility that thymic FSP1 + fibroblasts might regulate TECs through the released FSP1 protein. Three types of receptors, RAGE, annexin II and heparan sulfate proteoglycans were found to be surface receptors for FSP1 45 . We found that TECs expressed FSP1 receptors mainly including annexin II and syndecan1, further indicating that FSP1 could directly regulate TEC function. It is true that FSP1 supports TEC proliferation and maturation as evidenced by the in vitro and in vivo studies. FSP1KO mice had small thymus, low cellularity, decreased mTECs including CD40 + and CD80 + mTECs compared with WT mice and displayed inefficient regeneration after thymus damage. Addition of FSP1 protein significantly increased the cell number of TECs and mTECs, as well as the expression of CD80, CD40 and Aire expression on mTECs in FTOC and TEC culture assays. Furthermore, FSP1 significantly increased the expression of the key transcription factor for TECs, Foxn1 60,61 , and one of the important nursing receptors for TECs, FGFR2IIIb 53 , in mouse TEC culture system. Thus, FSP1 itself acted as a key direct regulator in TEC development through up-regulation of Foxn1 and FGFR2IIIb.
Taken together, our studies reveal a unique subpopulation of thymic fibroblasts expressing FSP1, which was mainly located in the thymic medullary zone. Thymic FSP1 + fibroblast subset plays an essential role in thymic medullary maintenance and regeneration under physiological and pathological situations. Thymic FSP1 + fibroblasts regulate the proliferation and differentiation of mTECs through providing IL-6, FGF7 and FSP1.

Materials and Methods
Mice. C57BL and streptavidin-PE (BD PharMingen). Control slides were incubated with isotype-matched Ig. Images were acquired with a two-photon microscopy (Carl Zeiss, Inc.).
Thymic stromal cell isolation and in vitro culture of TECs and thymic fibroblasts. Thymic stromal cells from the postnatal thymus were isolated as previously described 40,63 . In Brief, freshly dissected thymi were cut into 1-mm 3 pieces, washed with DMEM medium with 2% FBS several times to remove the majority of thymocytes. The thymic fragments were then incubated at 37 °C for 10 min in 2 mL solution of 1 mg/mL collagenase D (equivalent to 0.1% w/v) with 20 U/mL DNAse I (both from Sigma). Enzymatic treatment was repeated 3 times (the final incubation with collagenase/dispase enzyme mixture) until all fragments were dispersed. Gentle agitation was performed periodically at mid-and end-points of each digestion. Cell suspensions from each digestion were pooled in PBS containing 1% FBS and 5 mM EDTA to neutralize digestion and remove cell aggregates. Cells were centrifuged, re-suspended and filtered to remove clumps. Phenotypes of TECs were analyzed by surface FACS staining.
For TEC cell culture, thymi from WT neonatal mice were digested as mentioned above. Small thymic fragments from each step were collected and pooled. Fragments were allowed to settle and washed twice with PCT medium (CnT07, CellnTEC). The remaining thymic explants were plated in 48-well plates with CnT07 medium and cultured at 37 °C and 5% CO 2 for several days, during which TECs outgrew other stromal cells 40 . To determine the effect of FSP1 on TECs, cultured TECs were treated with 2 μ g/ml FSP1 protein. After 5 days treatment, TECs were collected with trypsin (Sigma) digestion and analyzed for MHCII and CD40 expression by FACS.
For thymic fibroblast culture, thymi from FSP1-GFP neonatal mice were digested as mentioned above. Cell suspensions from each digestion were pooled, centrifuged and re-suspended in the DMEM medium with 10% FBS and cultured for several days during which fibroblasts outgrew other stromal cells. To inhibit cell proliferation, thymic fibroblasts were treated with 10 μ g/ml mitomycin C for 4 hours at 37 °C and 5% CO 2 . FSP1 + cell deletion and thymus regeneration models. In systemic ablation model, FSP1-TK + and control WT mice were injected i.p. with ganciclovir (GCV) at a dose of 20 mg/kg body weight twice a day for 18 consecutive days. FSP1-TK + , FSP1KO and control mice were injected i.p. with Cy at a dose of 100 mg/kg body weight per day for two consecutive days to induce thymus damage 40 . The first day after Cy treatment was considered as recovery day 1. For FSP1 + cells-deletion, TK + and TK − mice were treated with GCV two times every day for 14 days. Thymocytes and TECs of these mice were isolated at indicated time for analysis.
Quantitative RT-PCR and ELISA. RNA was purified from cultured primary thymic fibroblasts and sorted TECs. mRNA was prepared using RNeasy Mini Kit (Qiagen) and the cDNA library was generated with Reverse Transcription System Kit (promega), according to the manufacturer's protocol. qPCR was performed using SYBR Premix Ex Taq (Takara). Relative expression values for target genes normalized to HPRT were obtained. The primers used in the present study were listed in Table 1. To determine the protein level of IL-6, FGF7 and FSP1, cell culture supernatants from FSP1 − and FSP1 + fibroblasts were collected and measured separately with ELISA commercial kits according to the manufacturer's protocol (USCN Life Science Inc.).
Bone marrow chimeras. Bone marrow cells (BMCs) from 8-wk-old TK − and TK + mice were prepared, respectively. 1× 10 7 BMCs were injected into the tail vein of lethally irradiated 8-wk-old TK − and TK + recipients to set up full chimeras as described 40,64 . 8 wks after reconstitution, recipient mice were treated with GCV or with combination of Cy and GCV. Mice were sacrificed at the indicated time points, and thymocytes and TECs were isolated for analysis.
Statistical analysis. All data are presented as the mean+SD. Student's unpaired t test for comparison of means was used to compare groups. A P value less than 0.05 was considered to be statistically significant.