Stat3 Signaling Promotes Survival And Maintenance Of Medullary Thymic Epithelial Cells

Medullary thymic epithelial cells (mTECs) are essential for establishing central tolerance by expressing a diverse array of self-peptides that delete autoreactive thymocytes and/or divert thymocytes into the regulatory T cell lineage. Activation of the NFκB signaling pathway in mTEC precursors is indispensable for mTEC maturation and proliferation resulting in proper medullary region formation. Here we show that the Stat3-mediated signaling pathway also plays a key role in mTEC development and homeostasis. Expression of a constitutively active Stat3 transgene targeted to the mTEC compartment increases mTEC cellularity and bypasses the requirement for signals from positively selected thymocytes to drive medullary region formation. Conversely, conditional deletion of Stat3 disrupts medullary region architecture and reduces the number of mTECs. Stat3 signaling does not affect mTEC proliferation, but rather promotes survival of immature MHCIIloCD80lo mTEC precursors. In contrast to striking alterations in the mTEC compartment, neither enforced expression nor deletion of Stat3 affects cTEC cellularity or organization. These results demonstrate that in addition to the NFkB pathway, Stat3-mediated signals play an essential role in regulating mTEC cellularity and medullary region homeostasis.


Introduction
The thymus provides a unique stromal microenvironment that is indispensable for the development of T cells that are both self-restricted and self-tolerant. Thymic epithelial cells (TECs) comprise the major component of the three-dimensional thymic stromal network, which also includes fibroblasts, dendritic cells (DCs) and endothelial cells [1][2][3]. TECs produce cytokines and cell surface ligands that promote thymocyte survival, growth and differentiation as well as chemokines that direct thymocyte migration into distinct cortical and medullary zones that contain phenotypically and functionally distinct TEC subsets [1,2,4]. Cortical TECs (cTECs) express Delta-like 4 ligands that engage Notch receptors on immature CD4 -CD8double negative (DN) thymocytes to promote their maturation to the CD4 + CD8 + double positive (DP) stage [5,6]. In addition, cTECs present self-peptide/MHC complexes that instigate positive selection of DP thymocytes with low avidity αβTCRs. Positively selected thymocytes differentiate to the CD4 + CD8or CD4 -CD8 + single positive (SP) lineage and migrate into the medulla where they reside for several days and are subjected to negative selection before export to the periphery [7][8][9][10][11][12][13][14][15].
Medullary TECs (mTECs) express a wide range of otherwise tissue-restricted antigens (TRAs) due, in part, to expression of the nuclear protein AutoImmune REgulator (AIRE) [16,17]. TRAs may be presented to SP thymocytes directly by expression on mTECs or indirectly by transfer from mTECs to DCs [18][19][20][21][22]. In either case, the unique ability of mTECs to express TRAs is essential for deletion of SP thymocytes bearing TCRs with high avidity for self-peptides, thereby establishing central tolerance [7,23]. In addition, mTECs are required to support the development of Foxp3 + T regulatory cells (Tregs), which function in the periphery to restrain immune responsiveness of self-reactive and potentially autoimmune T cells [24,25]. The essential contribution of mTECs in preventing the emergence and/or function of self-reactive peripheral T cells is evident from the autoimmune manifestations that occur in the absence of a functional mTEC compartment [26][27][28][29].
Considering the indispensable role that mTECs play in establishing central tolerance, the mechanisms regulating their differentiation, organization and maintenance are of considerable interest. It is well established that crosstalk between thymocytes and TECs is required for development and expansion of both cell types [30]. During fetal thymus development, lymphoid tissue inducer cells and γδTCR + thymocytes provide signals that initiate mTEC differentiation to generate small medullary islets [31,32]. In the postnatal period, signals from SP thymocytes play a nonredundant role in expanding the mTEC compartment to establish an organized and functional medullary region [33,34]. When thymocyte maturation is arrested at the DP stage, as occurs in TCRα deficient and MHC class I/MHC class II double deficient mice, small, hypocellular medullary foci are produced [35,36]. Numerous investigations have demonstrated that SP thymocytes promote mTEC development via activation of the NFκB signaling pathway. SP thymocytes express tumor necrosis factor superfamily (TNFSF) ligands such as CD40 ligand (CD40L), receptor activator of NFκB ligand (RANKL), lymphotoxin-α (LTα) and LTβ that bind corresponding TNFSF receptors on mTECs to activate NFκB signaling. In the absence of TNFSF ligands and/or receptors, mTEC differentiation and organization are severely impaired resulting in multi-organ autoimmunity [26,28,32,[37][38][39][40][41]. Similarly, deletion of RelB, TNF receptor-associated factor 6 (TRAF6) and other components in the noncanonical or canonical NFκB pathways blocks mTEC differentiation leading to the emergence of autoimmune symptoms [27,29,[42][43][44]. Furthermore, a recent study showed that signals from SP thymocytes are required to overcome TRAF3 mediated inhibition of noncanonical NFκB signaling [45].
Although it is well established that NFκB signaling is essential for mTEC development and that multiple TNFSF receptor/ligand combinations are involved in mTEC differentiation and expansion, the nature and severity of medullary defects vary depending on the particular component of the NFκB pathway that is inactivated. These findings imply that additional signaling pathways may contribute to the regulation of mTEC development and maintenance. Here we report that Signal transducer and activator of transcription 3 (Stat3) is required for optimal medullary region formation and mTEC homeostasis. Stat3 is a transcription factor that regulates diverse cellular processes including survival, proliferation and differentiation in a cell type specific manner [46,47]. We used Stat3 gain-of-function (GOF) and loss-of-function (LOF) genetic models to interrogate the contribution of Stat3 signaling to mTEC development and homeostasis. The data show that targeting a constitutively active Stat3 transgene to the mTEC compartment increases mTEC cellularity and expands the thymic medulla. Conversely, conditional deletion of Stat3 in mTECs reduces mTEC cellularity and impairs medullary region formation. In addition, we show that Stat3 activation is required for mTEC survival, but not for proliferation and may also play a role in differentiation. The precisely contrasting outcomes imposed by enforcing or preventing Stat3-mediated signaling in mTECs reveal that Stat3 activation plays a key role in medullary region formation and mTEC maintenance.

Selective expansion of the medullary region in K5.Stat3C transgenic thymi
The constitutively active Stat3 transgene (K5.Stat3C) is highly expressed in the thymic medulla and in scattered cells throughout the cortex as shown by immunohistochemical (IHC) staining for the Flag epitope tag (Fig 1A). Western blot analysis showed elevated levels of total and phosphorylated Stat3 in extracts from CD45 negative, but not CD45 positive cells, confirming that K5.Stat3C transgene expression is restricted to thymic stromal cells (Fig 1B). Overall thymus cellularity was reduced by~60% in K5.Stat3C transgenic mice compared to nontransgenic (NTg) littermates, hereafter referred to as NTg controls (Fig 1C). Hematoxylin and eosin (H&E) stained sections revealed a striking expansion of the medullary region and corresponding reduction of the cortical region in K5.Stat3C transgenic thymi (Fig 1E and 1F). Quantitative analysis of Aperio scanned slides confirmed the increased medullary to cortical ratio ( Fig  1D). IHC staining revealed that the K5.Stat3C transgenic thymus contained typical cTEC and mTEC subsets. Thus, K8 hi K5 -K14 -UEA-1 -cTECs were present in the cortex, whereas the medulla contained both K8 lo K5 + K14 + UEA-1and K8 hi K5 -K14 -UEA-1 + mTEC subsets (Fig 1G  and 1H) [48,49]. In contrast to the well-demarcated corticomedullary junction (CMJ) in NTg controls, K14+ mTECs in K5.Stat3C thymi extended beyond the CMJ into the cortex demonstrating that medullary expansion was associated with disruption of the typically circumscribed boundary at the CMJ.
Expression of the K5.Stat3C transgene in RAG-2 -/mice generates medullary region formation in the absence of thymocyte-derived signals We considered that the thymus hypoplasia and increased medullary/cortical ratio could be a consequence of stress-induced DP thymocyte depletion particularly since K5.Stat3C transgenic mice develop psoriasis-like skin lesions [50]. To resolve this issue, we introduced the K5.Stat3C transgene into RAG-2 -/mice in which thymocyte development is arrested at the CD44 -CD25 + DN3 stage. Medullary region development is severely compromised in the RAG-2 -/thymus due to the absence of positively selected thymocytes, which normally drive mTEC expansion [38,39,51]. The small medullary foci in RAG-2 -/thymi were populated with sparse clusters of K5 + K14 + mTECs and rare UEA-1 binding cells (Fig 2A, 2B and 2C). In contrast, RAG-2 -/-; K5.Stat3C thymi contained large medullary regions containing abundant K5 + K14 + and UEA-1 binding mTEC subsets. Furthermore, the pattern of Flag epitope expression indicated that the K5.Stat3C transgene is highly expressed throughout the expanded medullary regions, as well as in scattered cortical cells ( Fig 2D). DAPI negative areas were apparent in the vicinity of the enlarged medullary regions in RAG-2 -/-;K5.Stat3C thymi. Co-staining for K14 and VE-cadherin (S1 Fig) revealed that most of these regions are blood vessels, in accord with the concept that thymic vasculature plays a role in medullary compartment organization [52]. However, a few K14 bounded cyst-like structures were also present, consistent with active expansion and remodeling of this compartment.
Flow cytometric analysis verified that thymocyte differentiation remains arrested at the DN3 stage in RAG-2 -/mice that express the K5.Stat3C transgene ( Fig 2E). These data show that the medullary region expansion observed in RAG-2 sufficient K5.Stat3C thymi is not a trivial consequence of stress-induced cortical thinning. Rather, the results indicate that constitutive activation of Stat3 in RAG-2 -/-TECs bypasses the requirement for signals from positively selected thymocytes and independently drives medullary region formation.
The number of mTECs is increased in K5.Stat3C thymi K5.Stat3C transgenic and NTg control thymi were enzymatically dissociated for flow cytometric analysis of TECs, which were identified as CD45 -MHCII + EpCAM + cells. Both the percentage and absolute number of total TECs were increased in K5.Stat3C compared to control thymi (Fig 3A and 3B). The mTEC and cTEC subsets were distinguished by UEA-1 binding  and Ly51 expression [53][54][55]. Despite a decrease in the frequency of UEA-1 -Ly51 + cTECs ( Fig  3A), the absolute number of cTECs was comparable in K5.Stat3C and control thymi (Fig 3B). In contrast, there was a 2-fold increase in the number of UEA-1 + Ly51 -mTECs in K5.Stat3C thymi. Expansion of the mTEC, but not the cTEC compartment, is consistent with the increased medullary/cortical ratio observed in H&E stained sections of K5.Stat3C thymi ( Fig  1). However, the selective medullary expansion was not due to mTEC restricted transgene expression. Consistent with IHC stains showing Flag epitope positive cells in the cortex and medulla (Fig 1A), qRT-PCR analysis demonstrated that the Stat3C transgene is expressed in cTECs and mTECs (S2A Fig). Similarly, FACS analysis revealed increased levels of the active phosphorylated form of Stat3 protein in transgenic cTECs and mTECs compared to their NTg counterparts (S2B Fig).
We asked whether the expanded medullary regions and increased number of mTECs in K5. Stat3C thymi affected the cellularity of immature and mature thymocyte subsets. Flow cytometric analysis showed that transgenic and NTg control thymi contained comparable percentages of thymocytes in the major subsets defined by CD4 and CD8 expression, as well as in the DN subsets defined by CD44 and CD25 expression (Fig 3C and 3E). However, K5.Stat3C transgenic thymi contained fewer thymocytes in each subset including the T cell progenitor containing DN1 subset (Fig 3D and 3F). The factors responsible for decreased thymocyte cellularity are presently unknown. However, given the reduction in DN1 numbers, we speculate that indirect effects on availability and/or function of microenvironmental niches that support thymocyte progenitors could play a role [56,57]. Although cellularity and organization of the cTEC compartment appear normal, it is possible that defects in a minor cTEC component could account for the reduced thymocyte cellularity phenotype.
An immature MHCII lo CD80 lo mTEC subset is selectively expanded in K5.Stat3C thymi Previous studies have shown that functionally mature Aire positive mTECs expressing higher levels of MHC class II and CD80 arise from Aire negative precursors that express relatively low levels of MHC class II and CD80 [32,58,59]. Interestingly, there is a marked increase in the frequency and number of immature MHCII lo CD80 lo mTECs in K5.Stat3C compared to NTg control thymi (Fig 4A and 4B). Comparable results were obtained when CD40 rather than CD80 expression was assessed (S3 Fig).
That the expanded MHCII lo mTEC population does not express high levels of Aire is consistent with the notion that this subset consists of immature mTECs. However, Aire expression may be downregulated at later stages of TEC differentiation. A recent investigation of mTEC subset lineage relationships showed that Aire + CD80 hi mTECs can give rise to Aire negative mTECs in which CD80 is downregulated to intermediate expression levels [60]. Moreover, signals from positively selected thymocytes can induce Aire + mTECs to advance to an Airedifferentiation stage characterized by expression of involucrin, a marker of terminally differentiated keratinocytes [61]. Since Aire expression may be downregulated in differentiated mTECs, we asked if the expanded subset of MHC II lo CD80 lo Aire -mTECs in K5.Stat3C transgenic mice might represent a terminal stage of mTEC differentiation as opposed to an immature progenitor stage. We conclude that this is not the case because involucrin positive TECs were not increased in K5.Stat3C thymi ( Fig 4F). Moreover, the expanded subset of MHCII lo transgenic TECs expresses low to negative, rather than the intermediate levels of CD80 previously described for mature Aire negative TECs [60] (Fig 4A). These data indicate that the increased number of mTECs in K5.Stat3C thymi is a consequence of expansion of the immature MHCIIlo CD80 lo Aire -mTEC subset.
Similar to the thymic phenotype in RAG sufficient K5.Stat3C mice, the number of mTECs was significantly increased in RAG-2 -/-;K5.Stat3C thymi, primarily due to an~15 fold expansion of the MHCII lo mTEC subset (S4A Fig). Cellularity in the MHCII hi mTEC compartment also increased, albeit to a lesser extent (~6 fold) than that of the MHCII lo subset. Importantly, IHC staining (S4B Fig TRA expression is not altered and autoimmune symptoms are not detected in K5.Stat3C mice Expression of TRAs by mTECs is essential for negative selection of autoreactive thymocytes and establishment of central tolerance. Despite the lower percentage of Aire + mTECs ( Fig 4C) in K5.Stat3C thymi, there was no significant difference in the total number of transgenic MHCII hi Aire + mTECs compared to NTg littermate controls ( Fig 4D) and Aire expressing TECs were appropriately localized in the medulla (Fig 4E). Furthermore, qRT-PCR analysis of sorted MHCII hi mTECs verified comparable expression of Aire mRNA as well as Aire-dependent and Aire-independent TRAs in K5.Stat3C and control thymi (S5 Fig). Dendritic cells also play an important role in deleting autoreactive thymocytes by cross-presenting TRAs [19,22]. Mature mTECs secrete XCL1, a chemokine that is essential for proper accumulation of thymic dendritic cells in the medulla [62]. The presence of CD11c + dendritic cells at the corticomedullary junction and throughout the medullary region of K5.Stat3C thymi ( Fig 4G) suggests that transgenic mTECs function normally to recruit dendritic cells. Overt autoimmune symptoms were not found in K5.Stat3C transgenic mice up to 9 months of age. Histological examination of tissue sections revealed no evidence of lymphocytic infiltrations in kidney, lacrimal gland, liver or pancreas (S5D Fig). The K5.Stat3C transgene enhances mTEC survival but not proliferation Stat3 activation regulates a diverse set of cellular processes, including proliferation and apoptosis, in a cell type specific manner [63]. Both mature and immature mTECs contain a proliferative compartment, although mature mTECs have a higher frequency of cycling cells [58,59,64]. We analyzed BrdU incorporation to determine if expansion of the MHCII lo CD80 lo subset in K5.Stat3C thymi is a function of increased proliferative activity. Consistent with previous  Stat3 Regulates Medullary Thymic Epithelial Cells reports, we found a higher percentage of cycling cells in MHCII hi compared to MHCII lo mTECs in control NTg thymi [58,59]. There was no increase in the frequency of BrdU positive cells in MHCII lo or MHCII hi mTEC subsets from K5.Stat3C thymi indicating that expansion of the immature MHCII lo subset is not attributable to increased proliferation (Fig 5A and 5B).
To assess survival, we determined the frequency of apoptotic cells based on expression of cleaved caspase-3. The percentage of cleaved caspase-3 positive MHCII hi mTECs was comparable in K5.Stat3C and control thymi ( Fig 5C). However, there was a marked reduction in the percentage of cleaved caspase-3 positive cells in MHCII lo mTECs from K5.Stat3C thymi suggesting that enhanced survival contributes to expansion of this subset (Fig 5C and 5D). This notion was supported by qRT-PCR analysis showing that compared to NTg controls, K5. Stat3C MHCII lo mTECs express significantly increased levels of the anti-apoptotic gene bcl-2, which is a downstream target of the Stat3 signaling pathway (Fig 5E) [65,66]. Increased bcl-2 expression was also found in the MHCII hi mTEC subset from K5.Stat3C compared to NTg thymi; however, statistical significance was not achieved due to greater variability (P = 0.065). Analogous results were obtained when levels of Bcl-2 protein were assessed by FACS analysis (Fig 5F). A clear increase in Bcl-2 was apparent in MHCII lo mTECs, whereas only a modest increase was observed in the MHCII hi subset.

Conditional deletion of Stat3 disrupts medullary region architecture
The results obtained in K5.Stat3C mice demonstrate that constitutive activation of Stat3 promotes the survival and accumulation of MHCII lo mTECs. However, these data do not demonstrate whether Stat3 activity is indispensable for mTEC survival and/or differentiation. An earlier report found that Stat3 is not essential for proper mTEC development and medullary formation [67]. In that investigation, a K5 promoter driven Cre transgene was introduced into Stat3 fl/fl or Stat3 fl/mice to delete Stat3 in K5 positive TECs. The investigators reported that conditional deletion of Stat3 generated a hypoplastic thymus phenotype that progressed with age and was associated with loss of DP thymocytes, cortical atrophy and misalignment of cTECs [67]. However, the mTEC compartment appeared morphologically normal.
Using the same strategy, we introduced a K5.Cre transgene into mice that were homozygous for the same floxed Stat3 allele to examine the effect of Stat3 deletion in mTECs. PCR analysis showed efficient deletion of floxed Stat3 alleles in total TECs as well as in sorted cTECs and mTEC subsets (S7A and S7B Fig). Furthermore, FACS analysis confirmed extensive depletion of Stat3 protein in cTEC and mTEC subsets from conditional knock out compared to control thymi (S7C Fig). In contrast to the earlier study [67], we did not observe either thymus hypoplasia nor depletion of DP thymocytes in 6-12 week old K5.Cre;Stat3 fl/fl mice, hereafter referred to as Stat3 conditional knock out (Stat3 CKO) mice. Instead, the number and percentage of immature and mature thymocytes were comparable in CKO and littermate controls (Figs 6A, 6B and S9). Quantitative analysis of Aperio scanned H&E stained sections showed no alteration in the cortical/medullary ratio of Stat3 CKO compared to control thymi (S9 Fig). Nevertheless, striking abnormalities in medullary architecture and mTEC cellularity were obvious in the Stat3 CKO thymi. For example, there was an increase in medullary islets ( Fig 6C) and small cortical foci were frequently embedded in the medulla (Fig 6D). The medullary regions in Stat3 CKO thymi appeared tattered and were sparsely populated by mTECs that were dispersed in irregular clusters, rather than uniformly distributed throughout the medulla as in control thymi. (Fig 6E and 6F. Despite the paucity of mTECs, the presence of both K5 + K14 + and UEA-1 binding subsets suggests that Stat3 is not absolutely essential for mTEC development. The presence of Aire+ mTECs further indicated that mTEC maturation was at least partially intact (Fig 6G). In contrast to the relative scarcity of mTECs, CD11c+ dendritic cells were abundant in Stat3 CKO medullary regions indicating that recruitment of these cells was not impaired (Fig 6H). Conditional Stat3 deletion reduces the number of immature and mature mTEC subsets Consistent with the IHC data above, flow cytometric analysis showed that the percentage and number of TECs is decreased in Stat3 CKO thymi (Fig 7A and 7B). Loss of Stat3 selectively affected the mTEC subset as the number of cTECs was not altered. In stark contrast to the marked expansion of immature MHC II lo CD80 lo mTECs found in K5.Stat3C transgenic thymi (Fig 3), this subset was depleted by~50% in Stat3 CKO thymi (Fig 7C and 7D). The number of mature MHCII hi CD80 hi mTECs was also significantly decreased, a finding that is consistent with the precursor-progeny relationship between MHCII lo and MHCII hi mTEC subsets [32,58,59]. Despite reduced numbers, the percentage of Aire positive mTECs in Stat3 CKO thymi was comparable to that in controls (Fig 7E and 7F). Moreover, Stat3 CKO mTECs maintain the ability to express TRAs since expression levels of Aire-dependent and Aire-independent TRAs were comparable in FACS sorted MHCII hi mTECs from Stat3 CKO and control thymi (S10 Fig). Histological examination of tissue sections from kidney, lacrimal gland, liver and pancreas revealed no evidence of lymphocytic infiltrations in Stat3 CKO mice up to 9 months of age (S10 Fig) Moreover, we found no evidence for the presence of auto-reactive antibodies after staining RAG2 -/tissue sections with CKO serum (S11 Fig). Conditional Stat3 deletion does not alter the frequency of proliferating or cleaved caspase 3 positive mTECs, but does increase macrophage accumulation in medullary regions The frequency of proliferating mTECs in Stat3 CKO thymi was determined by FACS analysis of BrdU incorporation. We observed a comparable percentage of BrdU positive cells in MHCII lo and MHCII hi mTEC subsets from Stat3 CKO and control thymi (Fig 8A and 8B). Therefore, reduced mTEC proliferation does not account for depletion of the mTEC compartment in Stat3 CKO mice. Given the reduced percentage of cleaved caspase 3 positive cells observed in MHCII lo mTECs from K5.Stat3C transgenic thymi (Fig 5), we speculated that the converse result would be obtained for MHCII lo mTECs from Stat3 CKO mice. However, we did not find a reproducible increase in cleaved caspase 3 containing MHCII lo mTECs from Stat3 CKO compared to control thymi (Fig 8C and 8D). Increased apoptosis is often difficult to detect as apoptotic cells are rapidly cleared from the thymus. Furthermore, fragile apoptotic TECs may be lost during the multiple rounds of enzymatic digestion required to obtain single cell suspensions for FACS analysis. Therefore, we analyzed CKO and control thymic sections for the presence of macrophages as a gauge of apoptotic activity. IHC staining using either Mac-1 or F4/80 antibody revealed an increase in macrophages within and adjacent to the disrupted medullary regions of Stat3 CKO compared to control thymi (Fig 8E). Quantifying the number of macrophages per unit area of medulla ( Fig 8F) confirmed an increased accumulation of macrophages in the medullary regions of CKO thymi. These data support the notion that the reduction of mTECs in Stat3 CKO thymi is linked to increased cell death.

Discussion
The reciprocal medullary phenotypes found in K5.Stat3C transgenic and Stat3 CKO mice demonstrate that Stat3 activation regulates mTEC development in the adult mouse thymus. Medullary region expansion and increased numbers of immature mTECs were present in K5.Stat3C transgenic thymi. Moreover, expression of the K5.Stat3C transgene in RAG2 -/mice bypassed the requirement for thymocyte-derived signals in generating medullary regions containing functional mTEC subsets. In striking opposition, normal medullary architecture is disrupted and mTEC cellularity is reduced in Stat3 CKO thymi. The converse phenotypes in these GOF and LOF experimental models reveal that Stat3-mediated signaling is required for optimal medullary region formation and mTEC maintenance.
The presence of phosphorylated Stat3 in immature MHCII lo CD80 lo mTECs of steady-state thymi from young adult controls suggests that Stat3 signaling may be involved in thymus homeostasis. These results are consistent with a previous report that also implicated Stat3 in TEC homeostasis [68]. Thymocyte specific depletion of TRPM7, an ion channel and kinase dual function protein, partially blocked T cell development at the DN to DP transition. The corresponding progressive reduction in medullary region size was associated with a decrease in mTECs containing phosphorylated Stat3 [68]. Although a direct cause/effect relationship was not shown, these data were interpreted as suggesting that Stat3 activation is required for maintenance of the mTEC compartment [68]. While this report supports our present findings, the results we obtained for Stat3 CKO thymi differ starkly from another previous report in which deletion of Stat3 in K5 expressing TECs was found to alter the cTEC but not the mTEC compartment [67]. The reasons for this discrepancy are not clear but may involve differences in genetic background or in housing conditions that exacerbated stress leading to cortical thinning in the earlier report (64). In any case, our findings are entirely consistent with the results in an accompanying report from Satoh et al. who also found that Foxn1-Cre mediated deletion of Stat3 impairs medullary organization and reduces mTEC, but not cTEC cellularity.
It is well established that noncanonical NFκB activation is essential for mTEC differentiation and proliferation (reviewed in [3]). RelB deficient mice have small medullary foci that lack UEA-1 and Aire positive mTECs as well as dendritic cells [29,69,70]. The resulting defect in negative selection leads to extensive multi-organ lymphocytic infiltrates and autoimmunity. Similar autoimmune symptoms occur in the absence of signaling components upstream of RelB, such as NFκB inducing kinase (NIK) or IκB kinase α (IKKα) [27,43]. Furthermore, the absence of mature mTECs in mice lacking TRAF6 demonstrates the importance of the canonical NFkB pathway [42]. The aberrant medullary phenotype in Stat3 CKO thymi is less severe than that caused by defects in the NFκB signaling pathway. The contracted mTEC compartment remaining in Stat3 CKO thymi likely accounts for the lack of overt autoimmune manifestations. Nevertheless, the present study and the accompanying report (Satoh et al.) clearly demonstrate that deleting Stat3 in TECs results in disorganized medullary regions and a severe reduction in mTEC cellularity. Thus, we conclude that, in addition to NFκB signaling, Stat3 activation plays an indispensable role in medullary region formation and mTEC homeostasis.
Whether the Stat3 and NFκB signaling pathways interface in mTECs may depend on the particular ligand/receptor pairs that are engaged. Multiple ligands including RANKL, CD40L and LTβ are upregulated on positively selected thymocytes and activate NFκB signaling upon interaction with corresponding receptors on mTECs [26,38,39,71,72] These ligand/receptor frequency of MHCII lo CD80 lo and MHCII hi CD80 hi mTECs in Stat3 CKO and control thymi (D) Bar graph shows number (mean ± SD) of MHCII lo and MHCII hi mTECs (n = 5 for each). (E) Representative FACS plots show frequency of MHCII hi Aire + mTECs. (F) Bar graph shows number (mean ± SD) of MHCII hi Aire + mTECs (n = 5). *P<0.05 and **P<0.005 (Student's paired t-test) doi:10.1371/journal.pgen.1005777.g007 pairs differ in the extent to which they promote mTEC differentiation and/or proliferation. The absence of RANK or RANKL impairs mTEC proliferation and differentiation. As a consequence, the number of Aire positive mTECs is reduced, central tolerance is compromised and autoimmunity ensues [28,32,38]. The absence of CD40 or CD40L minimally affects mTEC cellularity or differentiation [28,38,71]; however, in the absence of both RANKL and CD40, medullary architecture is profoundly disrupted and the thymus is almost completely devoid of mTECs [28]. Furthermore, adoptively transferred spleen cells from RANKL and CD40 double deficient mice produced more extensive cellular infiltrates in liver and kidney of athymic nude recipients than spleen cells from mice deficient only in RANKL [28]. These data suggest that signaling pathways downstream of RANK and CD40 cooperate to promote mTEC differentiation and proliferation. Although the synergy between RANK and CD40 induced signals might arise from activation of both the canonical and noncanonical NFκB pathways, Stat3 activation may also be involved. In this regard, CD40 engagement has been shown to activate Stat3 signaling in B cells and cholangiocytes [73,74]. Therefore, it is possible that CD40-mediated Stat3 activation is involved in the synergy between RANK and CD40 signaling in mTECs. TGFβ was recently shown to negatively regulate mTEC maturation [75]. It is possible that Stat3 activity affects mTECs by directly or indirectly modulating TGFβ expression or its downstream signaling pathways.
The increased number of mTECs in K5.Stat3C transgenic thymi is due to selective expansion of the MHCII lo CD80 lo Airesubset. The increase in MHC II lo CD80 lo mTECs is not a result of greater proliferation, but instead is associated with a lower frequency of cleaved caspase 3 positive cells suggesting that constitutive activation of Stat3 confers enhanced survival to this TEC subset. Furthermore, the MHC II lo CD80 lo Aire -TECs in K5.Stat3C thymi express low levels of CD80 and are involucrin negative, indicating that these cells do not represent a post-Aire differentiation stage and supporting the notion that they are immature mTEC precursors. Although MHC II lo CD80 lo Aire -mTEC cellularity was increased, there was not a significant change in the number of mature MHC II hi CD80 hi Aire + mTECs despite the fact that the transgene is expressed and increased levels of phosphorylated Stat3 are present in mature MHC II hi mTECs. Although a higher turnover rate may prevent accumulation of transgenic MHC II hi CD80 hi Aire + mTECs, this explanation seems unlikely given that there was not a higher frequency of proliferating or caspase positive cells in the MHC II hi CD80 hi subset. It is possible that a block in progenitor differentiation restricts the number of mature mTECs. However, the similar number of MHC II hi CD80 hi mTECs in K5.Stat3C transgenic and NTg control thymi demonstrates that at least some of the MHC II lo CD80 lo mTEC subset in K5.Stat3C thymi can differentiate to the MHC II hi CD80 hi stage. We speculate that limited availability of microenvironmental niches in K5.Stat3C thymi restricts the number of mTEC progenitors that can complete maturation to the MHC II hi CD80 hi mTEC stage. Although further investigation is required to clarify the role of Stat3 signaling in mTEC differentiation, the presence of MHCII hi Aire positive mTECs in the expanded medullary regions of RAG-2 -/-;K5.Stat3C mice is consistent with the possibility that Stat3 activation may play a role in mTEC differentiation. Collectively, the data from this investigation and the accompanying report (Satoh et al.) demonstrate that Stat3 activation promotes mTEC survival and is essential for optimal medullary region development and homeostasis. Thus, in addition to the well-characterized requirement for canonical and noncanonical NFκB mediated signaling pathways, Stat3-mediated signaling contributes to optimal development and maintenance of the thymic medullary compartment.

Materials and Methods
Ethics statement MD Anderson Cancer Center IACUC approved the research performed with mice. The approval number is 00001097-RN00. MD Anderson is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALC). The AAALC number is 000183.
After termination of the experiments, animals were euthanized in a CO 2 chamber.

Mice
K5.Stat3C transgenic mice and K5.Cre;Stat3 fl/fl conditional knockout mice were generated and characterized as described previously [50,76,77]. Stat3 fl/fl mice and Stat3 fl/wt mice were littermate controls. Mice were maintained at the MD Anderson Cancer Center Department of Molecular Carcinogenesis in accordance with the guidelines set forth by the Association for the Accreditation of Laboratory Animal Care.

Morphometric analysis
Sections from 6-10 week old thymi were obtained from the center of the thymus over a 200 um distance, discarding 50 um between each depth. Ten sections were analyzed from each individual with 2 sections taken at each depth to provide a reliable representation of each thymus. Sections were stained with hematoxylin-eosin and scanned using Aperio Scanscope hardware. The digital copies of whole sections were saved and processed with Image Scope software from Aperio. The medullary and cortical areas were normalized against the area of the entire thymic section. A medulla/cortex classifier algorithm was employed to calculate relative medullary and cortical areas, expressed as the medulla/cortex ratio.

Western blot analysis
Protein lysates were made from FACS sorted CD45stromal cells and CD45 + hematopoietic cells using RIPA lysis buffer. The supernatant was separated by electrophoresis on 8-12% SDS-PAGE gels, electrophoretically transferred onto PVDF membranes and blocked with 5% non-fat dry milk in PBS with 0.1% Tween 20 for 1 hr at RT. Blots were incubated for 2 hr at RT with specific primary antibodies for Flag, phospho-Stat3, Stat3 and tubulin (Cell Signaling), were washed in PBS with 0.1% Tween 20 and incubated with HRP-conjugated secondary antibodies against rabbit or mouse (Bio-Rad). Blots were washed with PBS with 0.1% Tween 20 and detected with ECL Western Blotting kit (Pierce).

Quantitative real time PCR
Total RNA was purified from FACS sorted mTECs, cTECs and CD45 + hematopoietic cells and was purified using TRIzol (Life Technologies). cDNA was synthesized using a high capacity RT kit from Applied Biosystems following manufacturer's instructions. qPCR was performed with iTAQ SYBR Green Supermix (Bio-rad) or TaqMan master mix (Applied Biosystems). Reactions were run on ABI Prism 7900 HT and gene expression was normalized to endogenous αtubulin (Mm00846967_g1, Life Technologies). The mouse primer sequences used given in S1 Table. Supporting Information  Table. qRT-PCR primer sequences. (PDF)