Autoreactive T Cells and Chronic Fungal Infection Drive Esophageal Carcinogenesis

Ikkα^KA/KA Mice Develop ESCC Associated with Acquired Chronic Fungal Infection

We observed that C57BL/6 kinase-dead Ikkα^KA/KA (Zhu et al., 2007) developed esophageal epithelial hyperplasia, and approximately 20% of Ikkα^KA/KA mice at 5 months of age developed ESCC (Figure 1A). With increasing age, the morbidity of ESCC in Ikkα^KA/KA mice was increased. After 5 months, some Ikkα^KA/KA mice started to die due to developing systemic inflammation. These MESCCs expressed HESCC traits (Song et al., 2014; Stoner and Gupta, 2001), including a basal cell marker K5, significantly increased EGFR and STAT3 activities, elevated ΔNp63 expression, and downregulated tumor suppressor p16, p53, and IKKα expression (Figures 1A, 1B, and S1A). The K44A mutation destabilizes the IKKα protein in Ikkα^KA/KA mice (Figure 1B) (Xiao et al., 2013). IKKα regulates keratinocyte differentiation and proliferation independent of its kinase activity (Cao et al., 2001; Hu et al., 2001; Liu et al., 2008), indicating that a reduced amount of kinase-dead IKKα should not have a dominant negative activity in the esophageal tumorigenesis.

To determine a broad spectrum of molecular alterations that may drive the formation of MESCC, using an Array analysis, we found many molecular changes in MESCCs compared to wild-type (WT) esophagi (NCBI: GSE80005), which were also reported in HESCCs (Song et al., 2014) www.cbiportal.org) (Figure S1B and Table S1). These MESCCs highly expressed SCC hallmarks p63, K5/K14, and K6/K16, and expressed significantly increased or decreased levels of genes encoding proteins that regulate stem cell properties, cell cycle, DNA replication, epigenetic mechanisms, cell proliferation or survival, oxidative stress, DNA repair, inflammatory pathways, and PD-L1 (Figures 1C, S1C, and Table S2). Among these molecular alterations, EGFR, STAT3, Adams, ΔNp63, Smad3, and Trims are IKKα targets (Descargues et al., 2008; Liu et al., 2011; Liu et al., 2008).

In hematoxylin and eosin (H&E)–stained sections, we noticed fungus-like microorganisms on the surface of the esophagus in Ikkα^KA/KA mice, but not in WT mice (Figure 1D). The microorganisms were stained with Grocott’s methenamine silver (GMS), a stain used for fungi (Figure 1D). Oral swabs and ground esophagus extracts of Ikkα^KA/KA mice, but not WT mice, generated a high yield of fungal colonies in yeast peptone dextrose (YPD) agar plates containing the antibiotic chloramphenicol (Figure S1D). Sequencing of amplicons with ITS1-ITS4 and ITS1-ITS2 fungal-specific PCR primers allowed us to identify 19 different species of fungi from these colonies. These fungi belong to Ascomycota and Basidiomycota phyla (Figures 1E and S1E). C. cladosporioides and G. lagenaria were two major fungal species that colonized in the oral cavities and esophagi of Ikkα^KA/KA mice, suggesting that the fungi may have spread from the oral cavity to the esophagus. Using a direct sequencing approach, we detected an additional fungus, Fusarium sp. in Ikkα^KA/KA mice, but did not detect C. albican that was reported in human HESCC. Likely, it is because our mice and humans live in different environments. Cladosporium, Epicoccum, Candida (tropicalis), Alternaria, Aureobasidium, Penicillium, Fusarium, and Phoma genera were reported in human oral cavities or human cancers (Ghannoum et al., 2010; Underhill and Iliev, 2014). C. cladosporioides is one of the most common fungi both indoors and outdoors and can be pathogenic (Polack et al., 1976). In addition, we detected a low number of fungal colonies in the small intestine and colon of some Ikkα^KA/KA mice but none in WT mice (Figure S1F). We did not detect fungal colonies from the lungs and skin. Because we observed no tumors in the colon, we further investigated the activity of chronic fungal infection in esophageal tumorigenesis.

Ikkα^KA/+ and WT mice did not develop any phenotypes or chronic fungal infection after co-housing with Ikkα^KA/KA mice for more than 5 months. Ikkα^KA/KA female mice were unable to breed. Thus, Ikkα^KA/+ females were used to produce WT, Ikkα^KA/+, and Ikkα^KA/KA mice. We started to detect increased fungal infection in the oral cavities of Ikkα^KA/KA mice at 6–7 weeks of age. Likely, Ikkα^KA/KA mice acquired fungal infection from the environment.

Ikkα^KA/KA Mice Develop Impaired Central Tolerance and Systemic Inflammation

A marked infiltration of inflammatory cells was present in the pancreases, lungs, intestines, livers, kidneys, salivary glands, eyes, thyroids, and multiple stratified epithelial organs in Ikkα^KA/KA mice older than four to five months, compared to WT mice (Figures S2A and S2B, and data not shown). By using immunofluorescence (IF) staining, the serum from Ikkα^KA/KA mice, but not from WT mice, reacted with the kidneys, salivary glands, and pancreases of Rag1^−/− mice (Figure S2C), indicating that Ikkα^KA/KA mice developed systemic inflammation and autoantibodies.

We then hypothesized that the autoimmune phenotype of Ikkα^KA/KA mice may be secondary to a defect in central tolerance development (Lomada et al., 2007). We found a markedly decreased extension of the lightly stained medullary region that was filled with an increased number of small lymphocytes in H&E–stained sections of Ikkα^KA/KA thymuses, compared to WT (Figure 2A). IF staining verified that, compared to WT, Ikkα^KA/KA medullary region had a reduced number of mTECs as identified by K5, Aire, and lectin Ulex europaeus agglutinin-1 (UEA-1)–binding cells (Figure 2A). The intensity of Aire staining was reduced in Ikkα^KA/KA cells compared to WT. A frequency of CD11c⁺ dendritic cells was decreased in the Ikkα^KA/KA thymus compared to the WT (Figure S2D). The number of CD4⁺CD25⁺Foxp3⁺ regulatory T cells (Tregs) was reduced in Ikkα^KA/KA thymuses and spleens compared to WT (Figure 2B), which was presumably the result of a defect in negative T-cell selection. Thus, central tolerance development is impaired in Ikkα^KA/KA mice.

mTEC Development is Essential to Preventing Fungal Infection

To determine the molecular mechanism behind IKKα regulation of mTEC development, we analyzed gene-expression profiles. We found a downregulation in expression of genes encoding regulators for central tolerance development (Hamazaki et al., 2007; Lomada et al., 2007), including Aire; Aire’s target, TRAs: salivary protein 1, C-reactive protein, and fatty acid–binding protein; mTEC-expressed molecules: CD86, claudin3, and claudin4; several keratins; and molecules involved in cell survival and mitogenic pathways in Ikkα^KA/KA mTECs compared to WT (Figure 2C, NCBI: GSE52137). Using RT-PCR, we verified the expression of these genes in Ikkα^KA/KA and WT mTECs (Figure S2E). These molecular alterations may contribute to the defect in mTEC development in Ikkα^KA/KA mice.

Both canonical and noncanonical NF-κB activities are required for mTEC development and Aire expression (Akiyama et al., 2008). Whether IKKα regulates the canonical NF-κB pathway in mTECs was unknown. Treatment with an anti-CD40 antibody (an agonist) increased nuclear p65/p50 levels in the mTECs of WT mice, but not in Ikkα^KA/KA mice (Figure 2D), indicating a defect in the CD40-induced canonical NF-κB pathway in Ikkα^KA/KA mTECs. IKKα expression was reduced in Ikkα^KA/KA mTECs compared to WT. Expression of IKKβ and IKKγ in WT and Ikkα^KA/KA cells was similar. Treatment of mice with an anti-LTβR antibody (an agonist) induced an increase in nuclear p52 levels but did not alter p100 levels in WT cells. In contrast, in Ikkα^KA/KA cells, p52 and RelB levels were markedly decreased, and the agonist was not able to induce p52 nuclear translocation, causing increased p100 levels (Figures S2F and S2G). These results suggest that kinase-dead IKKα impairs nuclear translocation of both p50:p65 and p52:RelB NF-κB, resulting in a defect in mTEC development.

Keratin 5 (K5) is specifically expressed in mTECs. To verify the effect of epithelial-specific IKKα on NF-κB activities, mTEC development, and fungal infection, we introduced the K5.IKKα transgene, in which IKKα cDNA was overexpressed under the control of an epithelial cell–specific K5 promoter (Liu et al., 2008), into Ikkα^KA/KA mice. Unlike Ikkα^KA/KA cells, mTECs from Ikkα^KA/KA;K5.IKKα transgenic mice showed an increase in CD40-induced nuclear p50 and p65 translocation (Figure 2E). Also, K5.IKKα partially restored the size of the medullary regions and Treg numbers in the thymus, abolished the systemic inflammation and malignant phenotypes, and eliminated fungal infection in Ikkα^KA/KA;K5.IKKα mice (Figures 2F, 2G, S2H, and S2I). Ikkα^KA/KA;K5.IKKα mice lived for more than a year, maintaining an almost-normal phenotype. These results demonstrate that IKKα plays an essential role for activation of NF-κB pathways required for mTEC development. IKKα is also expressed in the basal squamous epithelial cells and its loss may contribute to the fungal growth on the mucosal surface of the esophageal epithelium and epithelial phenotypes.

TCR Vβ5.1 Repertoire Cells Are Increased in Ikkα^KA/KA Mice and K5.IKKα Suppresses the T-Cell Repertoire

Flow cytometric analysis and IF staining verified that K5.IKKα restored numbers of K5⁺UEA-1⁺ and CD45⁺integrinα6⁺UEA-1⁺ mTECs in Ikkα^KA/KA;K5.IKKα thymuses compared to Ikkα^KA/KA thymuses (Figures 3A and S3A). Expression of LTβR and CD80 was decreased in Ikkα^KA/KA mTECs compared to WT and K5.IKKα reversed this defect (Figure S3B), suggesting that K5.IKKα regulates the expression of the genes that are important for mTEC development. As mTECs determine the thymic microenvironment for selecting TCR repertoires (Akiyama et al., 2005), we examined the TCR Vβ repertoire including Vβ2, 3, 4, 5.1/5.2, 6, 7, 8.1/8.2, 8.3, 9, 10^b, 11, 12, 13, 14, and 17^a. The CD4⁺ and CD8⁺ T-cell populations expressing increased Vβ5 were detected in the Ikkα^KA/KA thymus and spleen compared to WT (Figures 3B–3D). Reintroduced K5.IKKα reduced Vβ5-positive T cells. We did not detect significant alterations in other repertoires. Because the Vβ5 antibody recognizes Vβ5.1 and Vβ5.2, we used PCR to verify increased Vβ5.1 and Vβ5.2 expression in T cells of Ikkα^KA/KA mice compared to WT and that K5.IKKα decreased Vβ5.1 and Vβ5.2 expression (Figures 3E and S3C). Increased Vβ5.1 repertoire was detected in the T cells from two sibling patients with APECED who carried AIRE mutations and fungal infection (Kogawa et al., 2002), suggesting that Ikkα^KA/KA T cells carry a signature of increased Vβ5.1 repertoire shared with APECED patients.

Most T cells found in Ikkα^KA/KA esophagi were CD4 cells (Figure S3D) and 85% of these T cells expressed activated- or memory-cell makers (CD3⁺CD44^Hi) in Ikkα^KA/KA esophagi compared to WT (Figure 3F). Notably, the TCR Vβ5 cell population was amplified in Ikkα^KA/KA esophagi compared to Ikkα^KA/KA thymuses and spleens (Figures 3B, 3C, 3G, S3E, and S3F). These results indicate that T cells in Ikkα^KA/KA esophagi are autoreactive.

To confirm whether the thymic microenvironment regulates T-cell development, we performed bone marrow (BM) transplantation using C57BL/6 Ikkα^KA/KA mice expressing a CD45.2⁺ marker and congenic C57BL/6 WT mice expressing a CD45.1⁺ marker. Irradiated Ikkα^KA/KA mice receiving Ikkα^KA/KA BM and irradiated WT mice receiving WT BM were used as controls. At 5 weeks after BM transplantation, the Vβ5 T-cell population was similar in the thymuses of chimeric Ikkα^KA/KA mice receiving WT BM or Ikkα^KA/KA BM and chimeric WT mice receiving WT BM or Ikkα^KA/KA BM (Figures 3H and S3G), indicating that the thymic microenvironment regulates TCR repertoire development. To examine whether this effect is induced by a T cell–intrinsic activity or not, we performed another transplantation and found that after irradiated Ikkα^KA/KA mice received mixed Ikkα^KA/KA and WT BM cells at a 1:1 ratio, the ratio of CD45.2⁺ and CD45.1⁺ cells from chimeric Ikkα^KA/KA mice remained similar (Figure 3I), demonstrating that Ikkα^KA/KA thymic microenvironment impairs T-cell development.

CD4 T-Cell Deletion or Antifungal Drug Treatment Prevents Fungal Infection and ESCC in Ikkα^KA/KA Mice

Impaired central tolerance mediates self-reactive T cell–associated autoimmunity. Next, we attempted to understand the role of autoreactive T cells in fungal infection, inflammation, and esophageal tumorigenesis. The numbers of infiltrating macrophages (F4/80) and T cells (CD3⁺) were markedly increased in Ikkα^KA/KA esophagi compared to WT (Figures 4A and S4A, top panel). All 20 Ikkα^KA/KA;Rag1^−/− mice that lack both T and B cells did not show esophageal epithelial phenotypes, oral fungal infection, systemic inflammation, and other organ phenotypes (Figures 4A and S4B, and data not shown). Similar results were obtained in all 20 Ikkα^KA/KA;Cd4^−/− mice, although these mice did show a few infiltrating inflammatory cells in some organs (Figures 4A, S4B, and S4C). The results suggest that Ikkα^KA/KA autoreactive T cells are associated with increased fungal infection, macrophage recruitment, and esophageal carcinogenesis. Then, we treated mice at 8 weeks of age with amphotericin B, an antifungal drug, for 3 months and monitored fungal infection by culturing oral swabs. The treatment significantly reduced the fungal burden, prevented macrophage infiltration, rescued esophageal phenotypes in 70% of Ikkα^KA/KA mice, and abolished esophageal tumors (Figures 4A–4C). The antifungal treatment resolved fungal infection better in the mouth than in the esophagus. The antifungal therapy did not rescue the mTEC defect in Ikkα^KA/KA mice (Figure 4A), suggesting that decreased fungal burden reduces inflammation and esophageal carcinogenesis.

Several studies reported increased cyclooxygenase-2 (COX-2) and PD-L1 expression in the vast majority of HESCCs (Chen et al., 2014; Zimmermann et al., 1999). Consistently, increases in COX-2 and PD-L1 expression as well as increased PD-L1–positive macrophages were detected in Ikkα^KA/KA esophagi compared to WT, and antifungal treatment reduced COX-2 and PD-L1 expression (Figures 4D, 4E, and S4A, bottom panel). We observed that the malignant epithelia expressing increased PD-L1 expression were tightly associated with infiltrating cells (Figure 4E). We did not see a marked increase in PD-L1 expression in other epithelial organs of Ikkα^KA/KA mice, suggesting that elevated PD-L1 expression in Ikkα^KA/KA esophagi is likely induced by the esophageal microenvironment. We detected an increased DNA damage marker, 8-OHdG, in Ikkα^KA/KA esophageal epithelia, and antifungal treatment reduced DNA damage (Figure 4F), suggesting that increased fungal burden increases DNA damage.

The absence of B cells in Ikkα^KA/KA;Igm^tm1Cgn mice did not rescue the inflammatory and malignant phenotypes, nor did it reduce fungal burden (Figures 4G, S4D, and S4E), suggesting that B cells are not a major cause for increased fungal burden and inflammation in the esophagi of Ikkα^KA/KA mice.

WT T Cells Eliminate Fungal Infection, Inflammation, and Esophageal Malignancy

To determine the effect of T cells on fungal infection, inflammation, and ESCC, we transferred WT and Ikkα^KA/KA T cells (5 × 10⁶ cells per injection for two injections) into 6 to 7-week-old Ikkα^KA/KA;Rag1^−/− mice. Lymphopenic mice do not develop increased fungal infections (Ostanin et al., 2007). Transferring Ikkα^KA/KA T cells but not WT T cells induced marked macrophage infiltration and fungal colonization on the esophagi of all Ikkα^KA/KA;Rag1^−/− mice (not in Rag1^−/− mice) at 4 months after the transfer. Two of 5 recipient Ikkα^KA/KA;Rag1^−/− mice developed esophageal malignancies (Figures 5A and S5A). Repeated experiments verified that the incidence of the esophageal malignancy in Ikkα^KA/KA;Rag1^−/− mice receiving Ikkα^KA/KA T cells was significantly higher than Ikkα^KA/KA;Rag1^−/− mice receiving WT T cells (p = 0.017, n = 10 mice/each group, Fisher’s exact test). Transferring Ikka^KA/KA or WT B cells (5 × 10⁶ cells per injection) did not induce severe phenotypes in 10 of 10 Ikkα^KA/KA;Rag1^−/− mice (Figure S5B). These results indicate that autoreactive Ikkα^KA/KA T cells permit fungal infection, resulting in inflammation and ESCC.

To test a therapeutic benefit of WT T cells, we injected WT or Ikkα^KA/KA T cells (5 × 10⁶ cells per injection) into 5-month-old and 10-week-old Ikkα^KA/KA mice twice during a two-week interval. At day 20 after the first injection, we found that WT T cells, but not Ikkα^KA/KA T cells, markedly reduced fungal colony numbers from oral swabs (Figure 5B). This inhibition of WT T cells was more efficient for young Ikkα^KA/KA mice than for old mice (Figure 5B), likely because the old mice had already developed a more severe fungal infection compared to the young mice. Histological examination consistently showed reduced numbers of infiltrating F4/80 and CD3 cells in the esophagi after WT T-cell therapy (Figure S5C and data not shown). These results suggest that WT T cells eliminate fungal burden and inflammation.

T cells produce IL17, IFNγ, and IL13, which can fight against fungi (Coste et al., 2008; Puel et al., 2011; van de Veerdonk et al., 2011). However, expression of these cytokines was lower in Ikkα^KA/KA T cells than in WT cells after treatment with an anti-CD3 antibody and zymosan A, a fungal extract (Figure 5C). The defect may contribute to initiating fungal colonization in Ikkα^KA/KA mice. Furthermore, Ifnγ, Il22, Il17a, and Il13 expression was partially rescued in the T cells of Ikkα^KA/KA;K5.IKKα mice compared to WT and Ikkα^KA/KA (Figure 5C), suggesting that epithelial-K5.IKKα influences the expression of these genes in Ikkα^KA/KA T cells. IKKα upregulates Il17a expression in T cells (Li et al., 2011). Thus, K5.IKKα may be unable to fully restore Il17a expression in Ikkα^KA/KA T cells.

Inflammation or EGFR Activity Increases Fungal Burden

We detected the increased expression of multiple cytokines and chemokines in Ikkα^KA/KA macrophages compared to WT (Figure S5D). Consistently, Ikkα^KA/KA esophagi expressed increased cytokines and chemokines compared to WT (Figure S5E). Antifungal treatment decreased inflammation in the esophagi of Ikka^KA/KA mice (Figures S5E, S1B, and S1C). We noticed an increase in Il17a expression in Ikkα^KAKA esophagi compared to WT but no increases in Il17a expression in Ikkα^KAKA T cells and macrophages (Figures 5C and S5D), suggesting that the esophageal epithelia may account for the increased Il17a expression. Thus, Ikka^KA/KA esophagi are inflamed with elevated cytokine/chemokine expression and macrophage numbers. Like APECED patients (Beerli et al., 2014), Ikkα^KA/KA mice developed increased anti-IL17F autoantibody (Figure S5F).

To examine the effect of inflammation on chronic fungal infection, we treated Ikkα^KA/KA mice at 3 to 8 months of ages with aspirin, a nonspecific anti-inflammation drug, for 20 days (Figure 5D). Aspirin treatment significantly reduced the number of fungal colonies in oral swabs. We used clodronate-loaded liposomes (Xiao et al., 2013) to deplete macrophages in Ikkα^KA/KA mice and obtained results similar to those from aspirin treatment (data not shown); however, liposome treatment caused the death of 5 treated Ikkα^KA/KA mice in a group containing 10 mice. These results suggest that inflammation contributes to the maintenance of chronic fungal infection.

It has been reported that fungi can interact with and activate EGFR and that treatment with an EGFR inhibitor significantly reduces the severity of oropharyngeal candidiasis (Zhu et al., 2012). EGFR is an IKKα target (Liu et al., 2008). We then treated Ikkα^KA/KA mice with GW2974, an EGFR inhibitor, for 30 days. The treatment significantly reduced fungal burden in oral swabs of Ikkα^KA/KA mice compared to untreated mice (Figure 5E). Thus, EGFR activation increased fungal burden.

Oral Fungal Infection Promotes Esophageal Tumorigenesis and Fungal Infection is Associated with HESCC

To determine whether increased fungal infection enhances tumorigenesis, we orally infected WT, Ikkα^KA/+, and Ikkα^KA/KA mice with Cladosporium cladosporioides (2 × 10⁷ fungi per mouse) every two weeks for a total of 4 infections. After infection, all Ikka^KA/KA mice, including untreated Ikka^KA/KA mice, maintained oral fungal infection. Fungal colonization in Ikka^KA/+ mice required more rounds of infections and WT mice were significantly resistant to fungal infection (Figures 6A and 6B). Ikkα^KA/KA mice started to die after the fourth infection. Histologic examination revealed that fungus-infected Ikkα^KA/KA mice developed increased numbers of esophageal or oral tumors that were K5 positive (Figures 6B, 6C, and S6A). These mice also developed hyperkeratosis and hyperplasia in the esophagi and oral cavities (Figure S6A and data not shown). Ikkα^KA/+ mice were normal. Infected Ikkα^KA/+ and WT mice developed hyperplasia in the esophagi and oral cavities, and the phenotype was more severe in Ikkα^KA/+ mice than in WT mice. We observed a robust correlation between increased esophageal phenotypes and increased fungal colonization among these mice. Oral tumors were detected in one of 16 infected Ikkα^KA/+ mice. Therefore, repeated fungal infections significantly promote the esophageal and oral tumorigenesis and IKKα protects the esophageal and oral epithelium from fungal infection and neoplastic transformation in a dose-dependent manner. We then propose a working model for mouse autoimmune disease, fungal infection, and MESCC development (Figure 6D).

HESCC is one of the most common and deadliest malignancies in China and South Asia. We also examined the association between fungal infection and non-APECED-derived HESCCs in a cohort containing 50 human normal esophageal tissue specimens and 80 stage-II and III HESCCs expressing K5 and p63 from Chinese non-APECED patients (Figures S6B, S6C, and S6D). We detected fungi in 78 of 80 (97.5%) human HESCC patients using a light microscope but not in normal human tissues (Figures 6E–6G). The association between fungal infection and HESCCs was statistically significant (Figure 6G), although the severity of the fungal infection was different among these HESCCs. We verified fungi in H&E–stained HESCC slides with Periodic acid-Schiff (PAS) and GMS staining and further confirmed these fungi in the paraffin sections by sequencing (Figure 6E). We extracted DNA from 3 H&E-stained HESCC slides and 2 normal esophagus slides, generated a PCR-product library for each case, and sequenced fungi. We identified 4 species of fungi from 3 HESCC patients, in which each HESCC case contained 2 or 3 different fungi: Candida albicans/Penicillium sp, Cladosporium cladosporioides/Penicillium sp, and Cladosporium cladosporioides/Ramularia mali/Penicillium sp, but not from controls (Figure 6E, right panel). C. albicans has been reported in HESCCs (Rautemaa et al., 2007). Most studies used specific culture media and morphological features to identify fungi in HESCCs (Mermel et al., 2009). The sequencing method we used may be more sensitive than other methods. Both Cladosporium and Penicillium are detected in human mouths (Ghannoum et al., 2010).

To determine shared hallmarks between HESCCs from non-APECED patients and MESCCs derived from Ikkα^KA/KA mice, we examined a HESCC tissue array containing 60 HESCCs and their adjacent tissues. We found increased numbers of macrophages (CD68) and T cells (CD3) and decreased IKKα levels in the HESCCs compared to the adjacent tissues (Figures 7A–7C and S7A). We often observed reduced IKKα expression in inflamed mouse organs. Thus, we hypothesized that inflammation may downregulate IKKα expression. Western blotting showed reduced IKKα expression in human SCC A431 and NCI 520 cell lines following treatment with TNFα or IFNγ at 72 or at 48 hr and did not show IKKα expression in human SCC-25 cells (Figures 7D and S7B), suggesting that increased cytokines downregulate IKKα expression in human SCC cells although the mechanism remains to be determined. The mutations of CHUK that encodes IKKα were detected in HESCCs (Figure 7E from www.cbiportal.org). Like MESCC, PD-L1 expression was significantly higher in the tumor regions than in the adjacent tissues (Figure 7F). Therefore, both HESCCs and MESCCs bear marked inflammatory cell infiltration, elevated PD-L1 expression, and reduced IKKα levels.

Chronic Fungal Infection Is an ESCC Promoter

Previously, the consequence of fungal infection in HESCC was not clear. A report (Rautemaa et al., 2007) showed that 6 of 92 APECED patients aged 29–44 years develop oral or esophageal SCCs. These cancer patients at infancy or in the early teenage years prior to SCC formation acquire chronic fungal infection. Like APECED, Ikkα^KA/KA mice developed chronic fungal infection prior to ESCC. Oral fungal administration promoted whereas antifungal treatment inhibited ESCC development. Therefore, chronic fungal infection promotes esophageal carcinogenesis.

C. albican is frequently reported in HESCCs (Rautemaa et al., 2007), although most studies lack sequence data. We did not detect C. albican in Ikkα^KA/KA mice. The mice used for this study are maintained in a specific pathogen–free (SPF) facility. These discrepancies are possibly due to different environments where people and mice live or due to a genotype. The effect of C. albican and other fungi on ESCC development needs to be tested.

The microflora of Ikkα^KA/KA mice did not transfer phenotypes to Ikkα^KA/+ and WT mice, indicating that a genotype determines this disease. Our unpublished data showed that treatment with antibiotics caused severe rectal prolapse with bleeding in Ikkα^KA/KA mice. This clinical problem is likely associated with mouse autoimmune conditions. The fecal flora is similar but the oral flora is distinct between Ikkα^KA/KA and WT mice. The bacterial burden is positively correlated with the fungal burden in the mouths of Ikkα^KA/KA mice. The detailed relationship between bacteria and fungi in carcinogenesis will be investigated in the future.

Critical Factors That Foster Chronic Fungal Infection

Enormous studies have provided insight into a cause of increased fungal infection by defects in immune system or cytokine products (Puel et al., 2011; Underhill and Iliev, 2014). However, the conditions that recapitulated fungus-associated ESCC development were previously unknown. In this study, we demonstrated that an impaired thymic microenvironment generated autoreactive T cells that permitted chronic fungal infection and incited tissue injury and inflammation. EGFR activation is associated with oropharyngeal candidiasis (Zhu et al., 2012). EGFR, an IKKα target, is an oncogene that can drive SCC development (Liu et al., 2008). Furthermore, we showed that autoreactive T cells alone were not sufficient to induce epithelial phenotypes, for example, Ikkα^KA/KA T cells induced phenotypes in Ikkα^KA/KA;Rag1^−/− mice but not in Rag1^−/− mice. Treatment with aspirin or an EGFR inhibitor reduced fungal burden in Ikkα^KA/KA mice. Macrophages activate EGFR (Xiao et al., 2013). Therefore, increased EGFR activity and inflammation provide the missing pieces for the establishment of chronic fungal infection in the epithelial organs. The defects in immunity and epithelial cells together facilitate and promote the development of chronic fungal infection and tumorigenesis. APECED is a complicated syndrome with diverse clinical manifestations. Aire deletion alone may not be sufficient to cause defects in immunity and oncogenes/tumor suppressors for generating fungal infection associated-malignant phenotypes in Aire^−/− mice. By contrast, IKKα expressed in mTECs and epithelial cells plays a crucial role for preventing fungal infection, autoimmunity, and epithelial phenotypes.

MESCC versus HESCC

APECED is a rare disease; thus, HESCCs from APECED are even more rare (Rautemaa et al., 2007). However, this fact does not reduce the importance of fungal infection in the development of HESCCs from non-autoimmune patients. Most HESCCs are frequently found in China and South Asia (Song et al., 2014), where the weather is moist, humid, and damp, which promotes fungi to thrive. Thus, people who live in these areas may frequently encounter fungi from the environment. Our study revealed a very high frequency of fungal infection in these HESCCs with different degrees of fungal burden. Some patients had a long history of smoking, which induces DNA damage and inflammation. Fungal infection induces Tnfα, Ifnγ, and Il17a expression in the monocytes and T cells of health people and APECED patients although Il17a expression in APECED is impaired (Ryan et al., 2008). We showed that TNFα and IFNγ downregulated IKKα expression. IKKα reduction, was frequently detected in HESCCs. IKKα haplodeficiency promotes carcinogen-induced carcinogenesis (Park et al., 2007). Therefore, fungal infection-induced inflammation and DNA damage can promote IKKα reduction-induced tumorigenesis in the esophagi. Notably, we detected increased T cell and macrophage infiltration and elevated PD-L1 expression in both MESCCs and HESCCs. Although normal T cells fight against fungal infection and tumor cells, PD-L1 blocks the antitumor activity of T cells. How this fungus-involved microenvironment enhances PD-L1 expression remains to be investigated. Clearly, fungus and EGFR can be used for therapeutic targets for HESCC treatment. We are aware of the toxicity of some antifungal drugs after long periods of treatment in humans. Furthermore, IL17A, which can be induced by fungi, plays a role in eliminating fungal infection (Conti et al., 2009). An increased IL17A amount is also pathogenic for human psoriasis (Gordon et al., 2016). Thus, the role of IL17A in ESCCs needs to be tested.

KEY RESOURCES TABLE

CONTACT FOR REAGENTS AND RESOURCE SHARING

Request for further information and resources may be directed to Lead Contact Yinling Hu (huy2@mail.nih.gov).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

MOTHOD DETAILS

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical parameters, including the exact value of n that represents number of mice per group, dispersion and precision measures (mean±SEM), and statistical significance are reported in the Figures and Figure Legends. Data is judged to be statistically significant when p <0.05 by selected statistical analysis method, including two-tailed Student’s t test for comparison between two groups, one-way ANOVA for comparison in more than two groups, Fisher’s exact test or Chi-square test for contingency table analysis. In figures, asterisks denote statistical significance as calculated by selected statistical analysis method (*, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001). Statistical analysis was performed in GraphPad PRISM 6.

DATA RESOURCES

The microarray analyses for mouse esophageal tissues and tumors (accession number NCBI: GSE80005) and for thymic epithelial cells (accession number NCBI: GSE52137) have been deposited at the NCBI Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/).