Genetic Alterations of TRAF Proteins in Human Cancers

The tumor necrosis factor receptor (TNF-R)-associated factor (TRAF) family of cytoplasmic adaptor proteins regulate the signal transduction pathways of a variety of receptors, including the TNF-R superfamily, Toll-like receptors (TLRs), NOD-like receptors (NLRs), RIG-I-like receptors (RLRs), and cytokine receptors. TRAF-dependent signaling pathways participate in a diverse array of important cellular processes, including the survival, proliferation, differentiation, and activation of different cell types. Many of these TRAF-dependent signaling pathways have been implicated in cancer pathogenesis. Here we analyze the current evidence of genetic alterations of TRAF molecules available from The Cancer Genome Atlas (TCGA) and the Catalog of Somatic Mutations in Cancer (COSMIC) as well as the published literature, including copy number variations and mutation landscape of TRAFs in various human cancers. Such analyses reveal that both gain- and loss-of-function genetic alterations of different TRAF proteins are commonly present in a number of human cancers. These include pancreatic cancer, meningioma, breast cancer, prostate cancer, lung cancer, liver cancer, head and neck cancer, stomach cancer, colon cancer, bladder cancer, uterine cancer, melanoma, sarcoma, and B cell malignancies, among others. Furthermore, we summarize the key in vivo and in vitro evidence that demonstrates the causal roles of genetic alterations of TRAF proteins in tumorigenesis within different cell types and organs. Taken together, the information presented in this review provides a rationale for the development of therapeutic strategies to manipulate TRAF proteins or TRAF-dependent signaling pathways in different human cancers by precision medicine.

With the rapid progress made in next-generation deep sequencing technology and the tremendous efforts put forth on whole genome/exome/transcriptome sequencing and copy number variation (CNV) analyses of cancers at the post-genome era, it has become increasingly clear that genetic alterations of TRAF proteins are commonly present in various human cancers.
Here we analyze the current evidence of genetic alterations of TRAF molecules available from the Cancer Genome Atlas (TCGA) (5) and the Catalog of Somatic Mutations in Cancer (COSMIC) (6) as well as the published literature, including the landscape of genetic alterations and the map of recurrent mutations in TRAF molecules in different types of human cancers. Moreover, we summarize the key in vivo and in vitro evidence that demonstrates the causal roles of genetic alterations of TRAF proteins in tumorigenesis within different cell types and organs. Collectively, the information presented in this review identifies TRAF proteins and TRAF-dependent signaling pathways as important therapeutic targets in specific human cancers.

Overview and Map of Recurrent Mutations
To date, there are 139 different mutations of the TRAF1 gene detected in human cancers, comprising 80% (111/139) mutations that alter the protein sequence of TRAF1 and 20% (28/139) coding silent mutations ( Table 1). In the TRAF family, TRAF1 has the lowest count of recurrent mutations. Only 29% (32/111) of the coding-altering mutations of TRAF1 are recurrent and have been detected in at least two patients with various cancers. Almost all the recurrent mutations of TRAF1 are missense mutations (94%, 30/32) except one nonsense mutation (truncation) and one fusion (Table 1 and Figure 2). These recurrent mutations occurred across 24 different amino acids that are distributed in all the major domains of the TRAF1 protein ( Figure 3). Interestingly, missense mutations of two specific amino acids are detected in more than three patients: R70C or H in the linker between the Zinc finger and the coiled-coil domain, and M182I of the coiled-coil (also known as TRAF-N) domain of the TRAF1 protein (Figure 3). The R70 mutations are detected in 4 patients with stomach, colon, and colorectal cancers (TCGA) (11)(12)(13). M182I is documented in 4 patients with melanoma and chronic lymphocytic leukemia (CLL) (14,15). The functional significance of R70C/H and M182I mutations of TRAF1 remains to be determined.

Fusion
There is only one fusion of the TRAF1 gene detected in human cancers, the TRAF1-ALK fusion that has been detected in five patients with anaplastic large cell lymphoma (ALCL) (16)(17)(18)(19). All five cases contain the identical in frame fusion of TRAF1 and ALK that generates a chimeric protein linking the Nterminal 1-294 aa of TRAF1 to the entire intracellular domain of ALK (1,058-1,620 aa), including its kinase domain (16)(17)(18)(19). Interestingly, expression of the TRAF1-ALK fusion protein leads to constitutive activation of the ALK and NF-κB pathways as demonstrated by the elevated levels of phosphorylated ALK (pALK) and STAT3 (pSTAT3) as well as nuclear p50 NF-κB1 and p52 NF-κB2 in ALCL cells (18). Similar to wild type (WT) TRAF1, the TRAF1-ALK fusion protein also binds to TRAF2 in co-immunoprecipitation experiments (18), suggesting the involvement of TRAF2 in the activation of NF-κB pathways. Furthermore, treatment of patient ALCL cells expressing the TRAF1-ALK fusion protein with proteasome inhibitors that decrease NF-κB1/2 or a selective ALK inhibitor (CEP28122) results in significant inhibition on lymphoma growth but could not eradicate lymphoma cells (18). Thus, constitutive activation of NF-κB1/2 pathways contributes to the neoplastic phenotype of TRAF1-ALK-expressing ALCL.

Landscape of Genetic Alterations
The frequency of genetic alterations of TRAF2 is generally <6% in human cancers ( Figure 1A) Figure 1B) (102). In addition, TRAF2 has been identified as an oncogene that is recurrently amplified and rearranged in 15% of human epithelial cancers ( Figure 1B) (103). Thus, the most common genetic alterations of TRAF2 are deep deletion, gene amplification and mutation (Figure 1). Truncation and fusion of TRAF2 are relatively rare but also detected in human cancers (Figure 1). two cancer patients. Recurrent mutations of TRAF2 are more complex than those of TRAF1, including not only missense mutations (82%, 75/92), but also frameshifts (11%, 10/92), truncations (4%, 4/92), in frame deletions (2%, 2/92) and fusion (1%, 1/92) ( Table 1 and Figure 2). TRAF2 recurrent mutations are identified at 52 different amino acids that are almost evenly distributed in all the structural motifs and domains of the TRAF2 protein (Figure 3). Interestingly, four mutation hotspots of TRAF2 are detected in more than 5 cancer patients, specifically P9, G10, R372, and Q457 (Figure 3). In particular, the frameshift deletion occurred at P9 (P9fs * 77) is found in 16 patients with colon cancer, colorectal cancer (CRC), uterine cancer, stomach cancer, and sarcoma, and an additional missense mutation at P9 (P9S) is also detected in a CRC patient (TCGA) (12,(104)(105)(106)(107)(108). The amino acid right next to P9, G10, also exhibits similar frameshift deletion (G10fs * 76) or insertion (G10fs * 70) or missense mutation (G10D) in five patients with colon cancer, CRC, gallbladder cancer, and glioblastoma (TCGA) (105,106,109). Missense mutations at R372 (R372C, H or S) of the TRAF-C domain of TRAF2 are detected in eight patients with HNSCC, melanoma, and prostate, uterine, cervical, stomach, and liver cancers (TCGA; COSMIC) (110)(111)(112)(113)  domain, TRAF-C domain, nuclear localization signals, and WD40 repeats. For each recurrent mutation, the nature of the mutation is indicated by a mutation symbol code and the patient count is indicated by a color code as shown at the bottom legend of the figure. The actual amino acid changes are also given for each recurrent mutation: letter change, missense mutation; *, Nonsense mutation (truncation); fs*, frameshift insertion or deletion; del, in frame deletion.

Fusion
There are five different fusions of the TRAF2 gene detected in human cancers, including TRAF2-CCDC183 in breast and bladder cancers, TRAF2-CACNA1B in bladder cancer, TMEM141-TRAF2 in breast cancer (TCGA), TRAF2-NOTCH1 in ovarian cancer (106) and NTRK2-TRAF2 in melanoma (115). Among these, only the TRAF2-CCDC183 fusion is recurrently detected in two patients with breast cancer and bladder cancer (TCGA). Functional contribution of these TRAF2 fusions to cancer pathogenesis is currently unclear.

In vivo Tumor Suppressive Roles
Inactivating mutations of TRAF2 are frequently detected in human MCL and DLBCL, resulting in elevated activation of NF-κB1 and NF-κB2 in malignant B cells (99)(100)(101)(102). TRAF2 is also involved in human MALT lymphomagenesis induced by the oncogenic cIAP2-MALT1 fusion protein through the interaction between TRAF2 and the BIR1 domain of cIAP2 portion of the fusion protein, leading to activation of the TRAF2-RIP1-NF-κB pathway (116). Consistent with the human evidence, B cell-specific TRAF2 −/− (B-TRAF2 −/− ) mice exhibit expanded B cell compartment in lymphoid organs due to constitutive NF-κB2 activation and survival advantage independent of the B cell survival factor BAFF ( Table 2) (117). Similarly in TRAF2DN-tg mice that express a dominant negative form of TRAF2 specifically in lymphocytes (Igh-TRAF2DN), inhibition of TRAF2 also leads to splenomegaly and lymphadenopathy due to constitutive NF-κB2 activation and increased numbers of B cells (40,42). Remarkably, TRAF2DN/Bcl-2 double-transgenic mice spontaneously develop small B cell lymphoma progressing to leukemia with many similarities to human CLL ( Table 2) (41,42). Thus, TRAF2 acts as a tumor suppressor in B lymphocytes primarily by inhibiting the NF-κB2 pathway through the wellestablished cIAP1/2-TRAF2-TRAF3-NIK axis (48,54,118). Genetic alterations of TRAF2 are detected in 1-2% of human liver cancers, including deletion, mutation and amplification (TCGA, PanCancer Atlas) (119). In human hepatocellular carcinoma (HCC), low expression of TRAF2 and its interacting partner RIP1 is associated with an unfavorable prognosis (43). In line with human evidence, deletion of both TRAF2 and RIP1 in liver parenchymal cells (LPC) leads to spontaneous development of hepatocellular carcinoma, which results from extensive hepatocyte apoptosis due to hyperactivation of caspase-8 but impaired NF-κB activation induced by TNFα ( Table 2) (43). Interestingly, TRAF2 also suppresses TNFα-induced necroptosis in hepatocytes by constitutively interacting with MLKL, thereby disrupting the TNFα-induced RIPK3-MLKL association and necroptosome formation. Induced TRAF2 deletion in adult mice results in rapid lethality, in conjunction with increased hepatic necroptosome assembly ( Table 2) (44). Therefore, TRAF2 protects hepatocytes from death and tumorigenesis by inhibiting both the TNFα-TNFR1-TRADD-FADD-caspase 8 apoptosis and TNFα-TNFR1-RIPK1-RIPK3-MLKL necroptosis pathways.
Genetic alterations of TRAF2 are detected in 3-4% of human HNSCC and melanoma ( Figure 1A). In cultured HNSCC cell lines, TRAF2 is required for cellular proliferation by acting in the TNFα-TNFR1-TRADD-TRAF2-RIPK1-TAK1-IKK-NF-κB pathway (120). In primary human keratinocytes, exposure to UV light triggers association of TRAF2 with TNF-R1 to induce NF-κB activation and inflammation (121). Keratinocytespecific TRAF2 −/− mice exhibit psoriatic skin inflammation associated with apoptotic death and epidermal hyperplasia, which is dependent on TNFα, constitutive NF-κB2 activation and inflammatory cytokine production (45). Further in support of a role for TRAF2 in skin tumorigenesis, mutations of the TRAF2-deubiquitinating enzyme CYLD are identified in patients with familial cylindromatosis, a condition that results in benign tumors of skin appendages, and CYLD −/− mice are highly susceptible to chemically induced skin tumors (122). Similarly, genetic alterations of TRAF2 are also identified in 2.7% (12/439) of human colon cancers (TCGA, PanCancer Atlas). In cultured primary human colon cancer cells, TRAF2 mediates the apoptosis by acting in the AMPK-ASK1-TRAF2-JNK-p53 axis in response to chemotherapies (123). Consistent with a potential role of TRAF2 in colon tumorigenesis, germline TRAF2 −/− mice spontaneously develop severe colitis, which results from TNFα-TNFR1-mediated apoptosis of TRAF2 −/− colonic epithelial cells and altered colonic microbiota (37). Interestingly, myeloid cellspecific ablation of TRAF2 markedly exacerbates DSS-induced colitis in mice due to enhanced TLR-induced proinflammatory cytokine expression in macrophages (46). This is caused by constitutively elevated levels of the transcription factors c-Rel and IRF5 that are targeted for proteasome-dependent degradation by the cIAP1/2-TRAF2-TRAF3 E3 ubiquitin ligase complex (46). Together, the above evidence consistently supports a suppressive role for TRAF2 in skin and colon tumorigenesis.
It is also noteworthy that genetic alterations of TRAF2 are detected in 2.6% (7/265) of human sarcomas (TCGA) and TRAF2 −/− mice display decreased viability of skeletal muscle tissue because of defective TNFα-induced NF-κB activation in myotubes ( Table 2) (38). Additionally, specific deletion of TRAF2 in T cells results in decreased numbers of CD8 naïve and memory T cells as well as NKT cells, due to impaired IL-15induced signaling in these cells ( Table 2). However, evidence of TRAF2 genetic alteration in T cell neoplasms is still lacking. Potential causal roles of TRAF2 dysregulation in muscle or T cell tumorigenesis remain to be elucidated.

Key Signaling Pathways in Cancer Pathogenesis
In addition to the above TRAF2-dependent tumor suppressive pathways verified in both human cancers and in vivo mouse models, several important tumor suppressive pathways involving TRAF2 have been suggested by evidence obtained from cultured human cancer cells or xenograft models. These are: (1) IRE1α-TRAF2-ASK1-JNK in the apoptosis of melanoma, lung cancer and OSCC cells induced by chemotherapies or ER stress (124)(125)(126); (2) the TRAF2-caspase-2 complex in mediating DNA damage-or chemotherapy-induced apoptosis of breast, cervical and lung cancer cells, in which TRAF2-mediated ubiquitination of caspase-2 stabilizes the caspase-2 dimer complex and enhances its activity to fully commit the cell to apoptosis (127,128); (3) TRAF2-mediated inhibition of constitutive NF-κB2 activation, cell proliferation, and anchorage-independent growth in pancreatic cancer, and a similar TRAF2-mediatied inhibition of the Eva1-induced NF-κB2-Sox2/CD15/CD49f pathway in the stemness of glioblastoma (129,130); and (4) TRAF2mediated K63-linked ubiquitination of MLST8 that disrupts the MLST8-SIN1-mTORC2-Akt pathway in the Kras-driven lung tumorigenesis (131). Together, these data suggest that TRAF2 is a tumor suppressor in many human cancers.
Interestingly however, increasing evidence indicates that TRAF2 also plays oncogenic roles in epithelial cancers and some Frontiers in Immunology | www.frontiersin.org Blunted airway inflammation and Th2 cytokine production in response to IL-25 administration due to defective IL-25R-Act1 signaling (66)

TRAF5
TRAF5 −/− Defective CD40-induced proliferation and surface molecule upregulation in B cells (67) Decreased CD40 plus IL-4-induced Ig production in B cells (67) Impaired CD27-induced survival and proliferation in CD4 and CD8 T cells (67,68) Defective GITR-induced proliferation, IL-2 production and NF-κB/p38/ERK1/2 activation in CD4 T cells (69) Enhanced OX40-induced Th2 differentiation of CD4 T cells and exacerbated Th2-driven lung inflammation (70) Enhanced IL-6-induced CD4 Th17 differentiation due to increased IL-6-gp130-STAT3 signaling and exaggerated Th17-driven experimental autoimmune encephalomyelitis Defects in generating CD8 memory T cells due to impaired AMPK-activation and mitochondrial fatty acid oxidation in response to growth factor withdrawal (84) Increased Th17 differentiation due to increased sensitivity of CD4 T cells to TGFβ-induced Smad2/3 activation and proliferation arrest (85) Impaired OX40-induced Th9 differentiation due to defective OX40-NIK-NF-κB2 signaling (86) Intestinal epithelial cell KO: TRAF6flox/flox, Villin-Cre Exacerbated DSS-induced colitis due to altered gut microbiota, which is independent of TLR signaling in intestinal epithelial cells (87) Skeletal muscle KO:TRAF6flox/flox, MCK-Cre Minimal muscle loss in response to transplanted tumor growth due to defective activation of NF-κB, ubiquitin-proteasome and autophagy-lysosomal systems (88) Improved regeneration of myofibers upon injury due to upregulated Notch signaling but downregulated NF-κB activation and inflammatory cytokine production Reduced starvation-induced skeletal muscle atrophy due to increased phosphorylation of Akt and FoxO3a and decreased AMPK activation (90) Direct evidence in tumorigenesis is highlighted in blue font.
Five mutation hotspots of TRAF3 are identified in more than 5 cancer patients, specifically N16, N285, K286, R310, and R376 (Figure 3). TRAF3 mutations at N16 have the highest patient count, including the missense mutation (N16T) identified in 10 patients with HNSCC (COSMIC) and the frameshift deletion (N16fs * 3) detected in a patient with splenic MZL (162,163). Mutations at the two consecutive amino acids N285 and K286 of the coiled-coil domain of TRAF3 exhibit the most complex pattern. N285 contains frameshift deletion (N285fs * 38), frameshift insertion (N285fs * 13) and missense mutation (N285S) identified in 8 patients with HNSCC, MZL, NPC, CRC, stomach cancer and uterine cancer (TCGA; COSMIC) (12,107,161,164,165). Similarly, K286 exhibits frameshift deletion (K286fs * 7 or fs * 11) and truncation (K286 * ) detected in six patients with B cell malignancies, including MM, CLL and WM (155,160,166,167). A third amino acid of the coiled-coil domain, R310, is consistently targeted by truncation (R310 * ) as detected in 8 patients with DLBCL, MM, HNSCC, cervical cancer and uterine cancer (TCGA) (113, 155,158,166,168). Missense mutations at R376 (R376W or Q) located in the linker between the coiled-coil and TRAF-C domains of TRAF3 are detected in six patients with lung cancer, CRC, SSCC, and melanoma (TCGA; COSMIC) (14,108,169). Many of these truncations, frameshifts and missense mutations have been shown to result in inactivation of TRAF3 by disrupting its interaction with NIK, thereby inducing constitutive NF-κB2 activation (155,156,158,159,170). Thus, most of the recurrent genetic alterations of TRAF3 identified in human cancers cause complete loss or inactivation of the TRAF3 protein.

In vivo Causal Roles in Cancer Pathogenesis
Similar to TRAF2 and also consistent with the frequent deletions and inactivating mutations of TRAF3 identified in human B cell malignancies ( Figure 1B), a tumor suppressive role for TRAF3 in B lymphocytes has been demonstrated by in vivo evidence obtained from mouse models. As shown for B-TRAF2 −/− mice, B cell-specific TRAF3 −/− (B-TRAF3 −/− ) mice also exhibit severe peripheral B cell hyperplasia due to prolonged survival of mature B cells independent of BAFF, which results from constitutive NF-κB2 activation (39,51). These B-TRAF3 −/− mice spontaneously develop splenic MZL or B1 lymphoma at high incidence (52). Interestingly, B-TRAF3 −/− mice also have doubled number of plasma cells due to enhanced responsiveness to IL-6 (171). Mechanistically, TRAF3 inhibits the IL-6-IL-6R-JAK1-STAT3 survival and differentiation pathway in plasma cells by facilitating the association of PTPN22 with JAK1 (171). Furthermore, the EBV-encoded oncoprotein LMP1 sequesters TRAF3 to produce functional TRAF3 deficiency in human and mouse B lymphoma cells (172,173). Intriguingly, lymphocytespecific TRAF3 transgenic mice also develop plasmacytosis, autoimmunity, inflammation, and cancers, which are likely caused by hyper-responsiveness of B cells to antigens and TLR agonists (55). Thus, TRAF3 acts as a tumor suppressor in naïve B cells, but an appropriate and balanced level, neither too high nor too low, of TRAF3 is required to maintain the homeostasis of plasma cells and protect them from tumorigenesis.
Interestingly, specific deletion of TRAF3 from myeloid cells (granulocytes, monocytes, and macrophages) leads to spontaneous development of histiocytic sarcomas derived from TRAF3 −/− tissue-resident macrophages in aging mice (56,174). The pathogenic mechanisms are likely related to the enhanced TLR-induced inflammatory responses observed in TRAF3 −/− macrophages through constitutive activation of NF-κB2, c-Rel, and IRF5, as described for TRAF2 −/− macrophages (56,174). Two other mouse models with functional relevance to TRAF3, Dok1 −/− Dok2 −/− Dok3 −/− mice and humanized TLR7/TLR8 transgenic mice, also spontaneously develop histiocytic sarcomas (175,176). DOK3, a negative regulator of protein tyrosine kinase (PTK)-mediated signaling, has recently been identified as a TRAF3-interacting protein (177). Similar to TRAF3 −/− macrophages, DOK3 −/− macrophages are defective in the TLR3-IRF3-IFNβ antiviral pathway (177). TRAF3 is also a transducer of TLR7 and TLR8 signaling through direct interaction with MyD88 (1). Transgenic expression of human TLR7/TLR8 in mice deficient for endogenous TLR7/TLR8 drives inflammation and proliferative histiocytosis, which can be reversed by compound deletion of MyD88 (176). Collectively, the above in vivo evidence indicates that TRAF3 is a tumor suppressor in macrophages and that dysregulation of the TLR-MyD88-TRAF3-Dok3 axis in macrophages plays causal roles in the pathogenesis of histiocytic sarcoma. However, because histiocytic sarcoma in humans is a rare malignancy with sparse pathologic and cytogenetic data (178,179), potential TRAF3 genetic alterations in human histiocytic sarcomas require future investigation.
In addition to the phenotype of histiocytic sarcoma, aging myeloid cell-specific TRAF3 −/− (M-TRAF3 −/− ) mice spontaneously develop chronic inflammation and other cancers that often affect multiple organs including the gastrointestinal tract (56). Similar to M-TRAF2 −/− mice, young adult M-TRAF3 −/− mice exhibit exacerbated DSS-induced colitis with increased levels of inflammatory cytokines produced by TRAF3 −/− macrophages in response to TLR signaling (46). Notably, another mouse model with functional relevance to TRAF3, NLRP12 −/− mice, is highly susceptible to colitis and colitis-associated colon cancer (180). NLRP12 interacts with both TRAF3 and NIK, and NLRP12 −/− cells have constitutively activated NF-κB2 associated with a decreased protein level of TRAF3 (180). Interestingly, both NLRP12 −/− hematopoietic and non-hematopoietic cells contribute to inflammation, but the latter dominantly contributes to colon tumorigenesis (180). In line with the in vivo data, mutations and deletions of TRAF3 are detected in 2.3% (10/439) of human colon cancers (TCGA, PanCancer Atlas). Furthermore, miR-32-TRAF3-mediated inhibition of the NIK-NF-κB2 pathway protects human colorectal epithelium against colorectal cancer in response to a diet of non-digestible carbohydrates (181). Thus, TRAF3 appears to act in both epithelial cells and myeloid cells to suppress colon tumorigenesis by inhibiting the NF-κB2 and TLR-induced inflammatory pathways.
Although most evidence identifies TRAF3 as a tumor suppressor, studies of the T cell-specific TRAF3 −/− (T-TRAF3 −/− ) mouse model suggest an oncogenic role for TRAF3 in T cells. Despite their constitutive NF-κB2 activation, TRAF3 −/− T cells exhibit impaired proliferation and activation in response to TCR and CD28 co-stimulation (39,57). T-TRAF3 −/− mice show defects in T cell-mediated immunity and IL-15-induced proliferation and survival of iNKT cells, and also have reduced number of CD8 central memory T cells ( Table 2) (57,59,60). Consistent with these in vivo data, TRAF3 is required for the proliferation of human neoplastic ALCL cells in culture (182). Silencing of TRAF3 in ALCL cells not only results in aberrant activation of the NIK-NF-κB2 pathway, but also affects the continued PI3K-AKT and JAK-STAT signaling (182). Therefore, distinct tumor suppressive and oncogenic roles of TRAF3 in different cellular contexts have been revealed from studies of both human and mouse models.

Key Tumor Suppressive Pathways
In addition to the above TRAF3-dependent tumor suppressive pathways verified in both human cancers and in vivo mouse models, several additional tumor suppressive pathways involving TRAF3 have been suggested by studies of cultured human cancer cells or xenograft models. These include: (1) TRAF3-mediated inhibition of the oncogenic RelB-SMAD4 association that represses TGFβ-SMAD target gene expression to promote the tumorigenesis of lung cancer, in which TRAF3 is targeted by RAS-NDP52-mediated autophagic degradation via the NDP52-TRAF3 interaction (183); (2) LIGHT-LTβR-TRAF3/TRAF5-ROS-ASK1-Caspase3 in the apoptosis of human colon cancer and hepatoma cells (184); (3) membrane-bound CD40L-CD40-TRAF3-p40phox-ROS-ASK1-MKK4-JNK-AP1caspase 9/3/Bax/Bak in the apoptosis of human bladder and CRC cells but not normal epithelial cells (185)(186)(187); and (4) TRAF3-mediated inhibition of the oncogenic RIP2-NF-κB1/NF-κB2/p38-Bcl-xL pathway in the survival and proliferation of glioblastoma cells (188). Taken together, available evidence supports that TRAF3 acts as a tumor suppressor in a variety of cell types, but we cannot rule out that TRAF3 upregulation might also alter normal cell homeostasis in the same or other cell types and therefore contribute to transformation, as it has been observed in B cells and T cells.

Overview and Map of Recurrent Mutations
There are 123 different mutations of TRAF4 detected in human cancers, comprising 85% (105/123) mutations that cause changes in the amino acid sequence of TRAF4 and 15% (18/123) coding silent mutations ( Table 1). About 42% (44/105) of the coding-altering mutations of the TRAF4 gene are recurrent and detected in at least two cancer patients, including mostly missense mutations (89%, 39/44), 3 truncations, 1 frameshift deletion, and 1 in frame deletion (Table 1 and Figure 2). TRAF4 recurrent mutations occurred at 32 different amino acids that are distributed in the entire length of the TRAF4 protein but are relatively enriched in the TRAF-C domain (Figure 3). Only two specific amino acids, R448 and R452 located at the C-terminal TRAF-C domain, are mutated in more than 3 patients (Figure 3). For R448, mixed missense mutations (R448Q or L) and a truncation (R448 * ) are identified in 4 patients with prostate cancer, uterine cancer, HNSCC, and OSCC (192)(193)(194)(195). For R452, missense mutations (R452W or Q or L) are detected in four patients with uterine, colorectal and lung cancers (10,108,193). Further studies are needed to determine whether such missense mutations in the TRAF-C domain result in loss-or gain-offunction of TRAF4.

Fusion
There is only one fusion of TRAF4 detected in human cancers, the TRAF4-FASN fusion identified in a glioma patient (TCGA), with currently unknown functional significance.

In vivo Causal Oncogenic Roles
Available human evidence indicates that gene amplification is the most common TRAF4 genetic alteration in cancers and that TRAF4 expression is ubiquitously elevated in many human cancers (196)(197)(198)(199)(200)(201)(202)(203)(204). This suggests that TRAF4 overexpression may play causal roles in cancer initiation, progression and metastasis. Similar to classical oncogenes (such as c-Myc and K-ras), TRAF4 is also required for ontogenic processes and TRAF4 −/− mice show defects in embryonic development and neurulation (61,62,205). Interestingly, TRAF4 −/− dendritic cells (DCs) derived from the null mice exhibit reduced in vivo and in vitro migration (64). Furthermore, recent in vivo evidence obtained from mouse models demonstrates the causal oncogenic roles of TRAF4 in skin tumorigenesis ( Table 2) (65). TRAF4 deficiency substantially diminishes IL-17A-induced ERK5 activation and epidermal hyperplasia in mice. In the DMBA/TPA-induced skin cancer model, TRAF4 −/− mice exhibit remarkably reduced tumor incidence and tumor numbers. Mechanistically, TRAF4 bridges the interaction between Act1 and MEKK3 in response to IL-17A signaling, and therefore is required for the activation of the downstream MEK5-ERK5-Steap4/p63 pathway. The transcription factor p63 directly induces TRAF4 expression in keratinocytes, promoting positive feedback on TRAF4 in the epidermis and thus sustaining the activation of the TRAF4-ERK5 axis to induce keratinocyte proliferation and skin tumorigenesis (65). These in vivo findings are reinforced by the examination of human SSCC samples, which also demonstrates the presence of the IL-17A-Act1-TRAF4-MEKK3-MEK5-ERK5-Steap4/p63 pathway (65). Together, both human and in vivo mouse evidence supports an oncogenic role for TRAF4.

Overview and Map of Recurrent Mutations
There are 188 different mutations of TRAF5 detected in human cancers, comprising 85% (160/188) mutations that alter the amino acid sequence of TRAF5 and 15% (28/188) coding silent mutations ( Table 1). Approximately 36% (57/160) of the coding-altering mutations of TRAF5 are recurrent in human cancers. Similar to TRAF4, TRAF5 recurrent mutations also include mostly missense mutations (85%, 49/57), but also some truncations (9%, 5/57), frameshift deletions (4%, 2/57), and an in frame deletion (2%, 1/57) (Table 1 and Figure 2). These recurrent mutations occurred at 36 different amino acids that are mainly enriched in the TRAF-C domain but also scattered in other regions of the TRAF5 protein (Figure 3). Mutations of three specific amino acids, R164, T232, and A548, are detected in more than three patients (Figure 3). Complex alterations of R164 of the zinc finger motif, including truncation (R164 * ) and missense mutations (R164Q or L), are detected in six patients with uterine, colon and bile duct cancers and DLBCL (TCGA) (12,193,215). Another missense mutation of the zinc finger motif, T232M, is detected in four patients with colon, breast, and prostate cancers (TCGA; COSMIC) (98). Missense mutation A548V of the TRAF-C domain is identified in four patients with uterine, cervical, stomach, and breast cancers (TCGA) (107). The functional consequences of these recurrent TRAF5 mutations await further investigation.

In vivo Causal Oncogenic Roles
Although not cataloged in TCGA, TRAF5 mutations are detected in 5% (5/101) of human DLBCL (102). TRAF5 expression is upregulated in human splenic MZL (216). In addition, apoptosisresistant B cell-acute lymphoblastic leukemia (B-ALL) cells have aberrantly higher protein level of TRAF5 and TRAF6 in response to irradiation than apoptosis-proficient B-ALL cells (217). The above evidence suggests that TRAF5 may be oncogenic in B cells. Consistent with human evidence, B cells of TRAF5 −/− mice show defects in CD40-induced proliferation and upregulation of surface molecules and activation markers as well as CD40 plus IL-4-induced Ig production (  (73). TRAF5 deficiency reverses the CD40-LMP1-induced enlargement of the spleen and lymph nodes, decreases the serum levels of IL-6 and autoantibodies that are elevated by CD40-LMP1-tg expression, and also inhibits LMP1-mediated JNK activation in B lymphocytes ( Table 2) (73). Together, both human and mouse evidence supports an oncogenic role for TRAF5 in B cells that appears to be required for LMP1-mediated B lymphomagenesis and is likely also involved in spontaneous B lymphomagenesis initiated by genetic alterations.
Additionally, available in vivo evidence indicates the importance of TRAF5 in the survival, proliferation and differentiation of different T cell subsets as detailed in Table 2, suggesting that TRAF5 malfunction may contribute to T cell malignancies. However, the evidence of TRAF5 alterations in human T cell lymphomas/leukemias is still lacking.  (224). The above evidence supports the hypothesis that TRAF5 plays oncogenic roles in various human cancer cells primarily by inducing NF-κB activation but also by activating the ERK1/2 pathway.

Overview and Map of Recurrent Mutations
There are 178 different mutations of TRAF6 detected in human cancers, comprising 85% (152/178) mutations that alter the protein sequence of TRAF6 and 15% (26/178) coding silent mutations ( Table 1). Only 27% (41/152) of the coding-altering mutations of TRAF6 are recurrently detected in at least two cancer patients. Similar to TRAF1, TRAF6 recurrent mutations also have the simplest composition and are almost exclusively missense mutations (93%, 38/41) except 2 truncations and 1 frameshift insertion (Table 1 and Figure 2). These recurrent mutations occurred at 30 different amino acids that are distributed in all the major domains but are enriched in the coiled-coil and TRAF-C domains of the TRAF6 protein (Figure 3). Mutations of only two specific amino acids, R335 and P398, are detected in more than three patients (Figure 3). A truncation (R335 * ) and missense mutation (R335Q) at R335 within the coiled-coil domain of TRAF6 are detected in five patients with colon and uterine cancers (TCGA) (12,234). Missense mutations at P398 (P398L or S) of the TRAF-C domain are identified in 4 patients with uterine, lung, and stomach cancers (TCGA) (8,193,235). Functional significance of these TRAF6 recurrent mutations in cancer pathogenesis remains to be elucidated.

In vivo Evidence of Potential Roles for TRAF6 in Tumorigenesis
Causal roles of TRAF6 in tumorigenesis have not been directly demonstrated with TRAF6 knockout or transgenic mice yet. However, available in vivo evidence supports potential contributions of TRAF6 dysregulation in tumorigenesis. Consistent with the genetic alterations (mainly amplification and mutation) and frequent overexpression of TRAF6 detected in human epithelial cancers such as breast cancer and uterine cancer ( Figure 1A) (229)(230)(231)(232)(233), deletion of TRAF6 in mice results in loss of NF-κB activity in epithelia and vasculature during mouse development ( Table 2) (78). Corroborating initial evidence, intestinal epithelial cell-specific TRAF6 −/− mice exhibit exacerbated DSS-induced colitis (87). In line with the in vivo data, knockdown of TRAF6 or inhibition of TRAF6 E3 ligase activity suppresses the survival, proliferation, migration, and metastasis of many human epithelial cancers, including breast, lung, liver, and colon cancers as well as HNSCC (230)(231)(232)(236)(237)(238)(239)(240). In the majority of these cases, the TRAF6-NF-κB axis is identified as the main oncogenic pathway, which is constitutively activated by TRAF6 overexpression or hyperactivated by upstream receptors such as TNFα, RANK, and TLR4/3 (236)(237)(238)(239)(240).
Similar findings have been obtained in the hematopoietic/lymphoid system. Hematopoietic-specific deletion of TRAF6 in mice leads to decreased tonic IKKβ-NF-κB activation, impaired hematopoietic stem cell (HSC) self-renewal and loss of hematopoietic stem/progenitor cells (HSPCs) in the bone marrow (BM) ( Table 2) (81). Knockdown of TRAF6 in human AML cell lines or patient samples results in rapid apoptosis and impaired malignant HSPC function as well as increased sensitivity to bortezomib (241). In the lymphoid lineage, TRAF6 mutations have been detected in 2.1% human DLBCL (TCGA) and 2.4% human cutaneous T cell lymphoma (CTCL) (242). TRAF6 −/− mice have reduced numbers of immature B cells in the BM and B cells from the null mice show defects in CD40 and LPSinduced proliferation and NF-κB activation ( Table 2) (74,75). B cell-specific TRAF6 −/− mice (B-TRAF6 −/− ) mice also exhibit reduced numbers of mature B cells in the BM and spleen as well as defective B1 B cell development ( Table 2) (82). Knockdown of TRAF6 or inhibition of the TRAF6-NF-κB axis induces apoptosis and cell cycle arrest in DLBCL, and also inhibits the proliferation and bone resorption of MM (243,244). Interestingly, T cell-specific TRAF6 −/− mice (T-TRAF6 −/− ) mice show multiorgan inflammation due to hyperactivation of the PI3K-Akt pathway in T cells ( Table 2) (83). T-TRAF6 −/− mice also exhibit increased Th17 differentiation due to enhanced sensitivity of CD4 T cells to TGFβ signaling (85), but have defects in generating CD8 memory T cells caused by defective AMPK activation in activated CD8 T cells (84). In addition, T-TRAF6 −/− mice exhibit impaired OX40-induced CD4 Th9 differentiation, which requires TRAF6mediated activation of the NIK-NF-κB2 pathway in CD4 T cells ( Table 2) (86). In support of a role for TRAF6 in T cell tumorigenesis, inhibition of the TRAF6-NF-κB-c-Myc axis through miR-146a/b-mediated downregulation of TRAF6 delays PTEN −/− T cell lymphomagenesis in mice (245). Furthermore, inhibition of the IRAK1/4-TRAF6 axis sensitizes human T cell ALL (T-ALL) to chemotherapies (246). Collectively, the above evidence is consistent with the hypothesis that TRAF6 may serve as an oncoprotein in epithelial cancers and hematopoietic/lymphoid neoplasms mainly through inducing aberrant NF-κB activation.
Interestingly, in vivo evidence also indicates the functional importance of TRAF6 in the brain and muscle ( Table 2). TRAF6 −/− mice show defective neural tube closure and exencephaly (80). Mechanistically, TRAF6 interacts with the p75 neurotrophin receptor (p75NTR), and thus is required for NGF-induced NF-κB activation and survival in Schwann cells as well as BDNF-induced JNK activation and apoptosis in superior cervical ganglia neurons (79). In skeletal muscle, TRAF6 deficiency prevents muscle loss and cancer cachexia in response to transplanted tumor growth, improves regeneration of myofibers upon injury and reduces skeletal muscle atrophy upon starvation through regulating NF-κB activation/ubiquitinproteasome/autophagy-lysosomal systems, Akt/FoxO3a/AMPK activation and Notch signaling, respectively (88)(89)(90). In line with the mouse data, genetic alterations of TRAF6, including amplification, mutation and deletion, are detected in 1% of human glioblastoma (247,248) and 3% of human sarcoma (TCGA) (9). It would be interesting to test potential causal roles of TRAF6 in brain and muscle tumorigenesis in future studies.
Intriguingly, several TRAF6-dependent tumor suppressive pathways have also been described for human cancers in the literature. Examples are: (1) TRAF6-p62-mediated inhibition of the HK2 glycolytic activity and the growth of liver cancer, in which TRAF6 catalyzes the K63-linked ubiquitination of HK2 and targets HK2 for p62-mediated autophagic degradation (275); (2) TRAF6-mediated suppression of the EZH2-H3K27me3 pathway and the progression of prostate cancer, in which TRAF6 mediates K63-linked ubiquitination and degradation of EZH2 (276); and (3) TRAF6-mediated decrease of the H3K4me3 level and thus the tumorigenesis of prostate cancer, in which TRAF6 mediates K63-linked ubiquitination of JARID1B to increase the demethylase activity of JARID1B on H3K4me3 (277). Taken together, both tumor suppressive and oncogenic roles of TRAF6 have been reported in human liver cancer and prostate cancer. As discussed for TRAF2, this phenomenon may be related to the mutational profile and malignant stage of the cancer cells as well as the nature of the environmental cue or treatment regimen.

Overview and Map of Recurrent Mutations
In the TRAF family, TRAF7 has the highest counts of total and recurrent mutations. There are 376 different mutations of TRAF7 detected in human cancers, including 87% (326/376) codingaltering mutations and 13% (50/376) coding silent mutations ( Table 1). Over half (53%, 174/326) of the TRAF7 coding-altering mutations are recurrently detected in at least two cancer patients. TRAF7 recurrent mutations are mostly missense mutations (92%, 161/174). Small percentages of other recurrent mutations include 5 frameshifts, 3 truncations, 2 in frame deletions, 2 splice mutations, and 1 fusion (Table 1 and Figure 2). These recurrent mutations occurred at 89 different amino acids covering different regions of the entire length but highly enriched in the last 4 WD40 repeats of the TRAF7 protein (Figure 3). Of particular interest, missense mutations of six specific amino acids located within the C-terminal WD40 repeats, N520, H521, G536, S561, K615, and R641, are identified as mutation hotspots of TRAF7 detected in more than 15 cancer patients (Figure 3). N520 mutations (N520S, H, Y, or T) are found in 31 patients with meningioma, mesothelioma, sarcoma and colon cancer (12,106,282,284,285,293,294). Mutations of the next amino acid H521 (H521R or N) are identified in 15 patients with adenomatoid tumor, perineurioma, and meningioma (281,283,284). G536 mutations (G536S or V) are detected in 16 patients with meningioma, pancreatic cancer, mesothelioma and stomach cancer (106,282,284,285,(293)(294)(295). S561 mutations (S561R, N or T) are identified in 19 patients with adenomatoid tumor, perineurioma and meningioma (281,283,284). K615E mutations are detected in 15 patients with meningioma and OSCC (284,296). R641 mutations (R641H, C, P, or L) are detected in 24 patients with uterine, bile duct, colon, stomach and lung cancers and meningioma (TCGA, PanCancer Atlas) (8,106,282,284,285,294,(297)(298)(299). Although the functional significance of most TRAF7 mutations is currently unclear, the exceptionally high recurrence and clustering of missense mutations implicate TRAF7 malfunction as a critical pathogenic event in relevant human cancers.

Fusion
There are six different fusions of TRAF7 and other genes detected in human cancers, including TRAF7-LRRC1 in lung cancer, GFER-TRAF7 in mesothelioma, CORO7-TRAF7 in glioma, TRAF7-MAPK8IP3 in bladder cancer, and TRAF7-RAB26 and E4F1-TRAF7 in ovarian cancer (TCGA) (106). Among these, only the TRAF7-LRRC1 fusion is recurrently detected in two patients with lung cancer (TCGA). However, all the TRAF7 gene fusions have not been verified at the protein level and their functional consequence is unknown.

In vivo Evidence of Potential Roles in Cancer Pathogenesis
No TRAF7 −/− or TRAF7-tg mouse model has been published yet. Importantly however, Tokita et al. recently reported that de novo missense mutations in TRAF7 cause developmental abnormalities and other clinical symptoms in seven unrelated patients, including developmental delay (5/5), congenital heart defects (6/7), limb and digital anomalies (7/7), and dysmorphic facial features (7/7) (300). TRAF7 mutations identified in this study include a recurrent R655Q mutation found in four patients, and another 3 single mutations each identified in one patient, including K346E, R371G, and T601A (300). Interestingly, R371 recurrent mutations are also detected in human cancers (Figure 3). K346 is a ubiquitination site of TRAF7 (301). Both K346 and R371 are located in the coiled-coil domain of TRAF7 that is important for TRAF7 homodimerization (302,303). The recurrent R655Q mutation has also been previously identified as a de novo event in an autism patient (304). Both T601 and R665 are located in the C-terminal WD40 repeats of TRAF7, which contain most mutation hotspots of TRAF7 detected in human cancers (Figure 3) and are known to mediate the interaction of TRAF7 with MEKK3 or c-Myb (302,305). Tokita et al. further revealed that transfection of the R665Q, T601A, or R371G mutants of TRAF7 into HEK293T cells results in significantly reduced levels of ERK1/2 phosphorylation, both basal and in response to TNFα signaling (300). Consistent with this biochemical evidence, conditional ERK2 −/− mice show a phenotype mirroring that observed in the seven patients with TRAF7 mutations, including craniofacial abnormalities, cardiovascular malformations and limb defects (306). These highly interesting findings warrant further investigation of the in vivo functions of TRAF7 mutations in cancer pathogenesis using animal models.

TRAF7-Mediated Signaling Pathways
Compared to other TRAF proteins, the signaling mechanisms of TRAF7 are understudied and remain poorly defined (289,307). In addition to the above TRAF7-ERK1/2 pathway revealed by studying TRAF7 mutants of patients with developmental defects, the following TRAF7 signaling pathways have been proposed based on in vitro studies. (1) Transfection of tumorderived TRAF7 mutants (H521R, Y538S, or S561R) but not WT TRAF7 in 293T cells causes increased phosphorylation of RelA and expression of the NF-κB target gene L1CAM, which is also elevated in adenomatoid tumors (281). (2) Overexpression of TRAF7 or TNFα induces caspase-dependent apoptosis in HEK293 and HeLa cells via the TRAF7-MEKK3-NF-κB/p38/JNK-AP1/CHOP pathway, in which TRAF7 interacts with MEKK3 and potentiates the kinase activity of MEKK3   (308). (4) TRAF7 represses TNFαinduced NF-κB activation to enhance apoptosis in HEK293 cells by promoting K29-linked ubiquitination and lysosomal degradation of NEMO and RelA (309). Paradoxically, TRAF7 is identified as an activator of the NEMO-RelA-NF-κB-cyclin D1 pathway in mouse myoblasts and thus a suppressor of myoblast differentiation (310). (5) TRAF7 participates in TLR2-induced production of inflammatory cytokines (TNFα, IL-1β, and IL-8) in A549 and HeLa cells by acting in the TLR2-TRAF6/TRAF7-IKK1/2/NEMO-NF-κB and TLR2-TRAF6/TRAF7-MKK3/6-p38 pathways (311). (6) TRAF7 inhibits the transcriptional activity of the oncoprotein c-Myb in M1 mouse leukemia cells and DND39 human Burkitt's B lymphoma cells by directly interacting with c-Myb and stimulating the sumoylation of c-Myb to sequester c-Myb in the cytoplasm (305). (7) TRAF7 mediates K48-linked ubiquitination of p53 as demonstrated by an in vitro ubiquitination assay, which likely induces p53 degradation. Correspondingly, TRAF7 protein is downregulated and p53 protein is upregulated in a panel of breast cancer specimens, and TRAF7 downregulation correlates with poor prognosis in breast cancer (312). In summary, WT TRAF7 appears to be a tumor suppressor that promotes cell apoptosis. TRAF7 mutations or downregulated protein levels may lead to aberrant NF-κB activation or altered signaling of ERK1/2, p38, JNK, c-FLIP, c-Myb, or p53 to drive tumorigenesis. Further studies are required to clarify the roles and mechanisms of TRAF7 alterations in cancer pathogenesis.

COMBINED GENETIC ALTERATIONS OF ALL TRAFS IN THE SAME HUMAN CANCER
After analyzing the genetic alterations of each TRAF gene in human cancers individually, we next examined the combined genetic alterations of all TRAFs in the same type of human cancer using the TCGA tool. Although the frequency of the genetic alterations of each TRAF is generally low (usually <5%), their combined rate is substantially increased to 10-35% in many types of human cancers (Figure 4) (98). It is interesting that genetic alterations of different TRAFs are often mutually exclusive in the same cancer patient and simultaneous genetic alterations of two or three different TRAFs in the same cancer patient are generally rare events (Figure 4) except for ovarian cancer (TCGA, Provisional), uterine cancer and melanoma (TCGA, PanCancer Atlas). We summarize key oncogenic pathways that involve multiple TRAF proteins in skin carcinogenesis as depicted in Figure 5 as well as in B cell malignancies as depicted in Figure 6, both of which have been verified by studies of human cancers and in vivo mouse models. However, we believe current understanding only represents "the tip of the iceberg" of oncogenic mechanisms involving TRAF proteins. Given the often mutually exclusive genetic alterations of different TRAFs in the same cancer, it is very likely that all seven TRAFs may have non-overlapping and distinct contributions to different aspects or at different stages of the initiation, progression and metastasis of the same cancer. These unanswered questions represent fascinating areas for future exploration.

TRAF PROTEINS IN PATHOGEN-INDUCED CARCINOGENESIS
The importance of TRAF proteins in cancer pathogenesis is strengthened by mounting evidence that demonstrates their involvement in pathogen-mediated carcinogenesis in certain human cancers. For example, chronic infection with the bacteria Helicobacter pylori (H. pylori) is a major cause of gastric cancer. H. pylori infection induces TRAF1 overexpression and the expression of the transcription factor Cdx2 in both human gastric epithelial cells and mice, which are mainly driven by NF-κB activation. TRAF1 overexpression plays an antiapoptotic role in H. pylori-infected gastric cells (316). Induction of Cdx2 contributes to intestinal metaplasia, a precursor event to gastric cancer. Interestingly, TRAF3 inhibits H. pylori infectioninduced NF-κB activation and Cdx2 expression, and is required to resist the infection by acting in the NOD1-RIP2-TRAF3 pathway in gastric epithelial cells (317,318). Furthermore, the oncoprotein cag PAI of H. pylori activates NF-κB and induces IL-8 secretion through the TRAF2/TRAF6-NIK-IKK pathway in gastric cancer cells (319). Another carcinogenic factor, Tip-α, of H. pylori activates NF-κB by inhibiting the expression of miR-3178, which directly targets TRAF3 mRNA for downregulation, in gastric epithelial cells (320). Therefore, H. pylori chronic infection-induced gastric tumorigenesis involves activation of TRAF1, TRAF2, and TRAF6 as well as inhibition of TRAF3 (Table 3).
A variety of viral infections have also been linked to cancer development. DNA viruses, such as Epstein-Barr virus (EBV), Kaposi's sarcoma-associated herpesvirus (KSHV), human cytomegalovirus (HCMV) and human papilloma virus (HPV), cause NPC, B lymphomas, breast cancer, glioma, cervical cancer, and HNSCC in the host (335,336). RNA viruses, such as hepatitis C virus (HCV) and human T-cell leukemia virus type 1 (HTLV-1), may lead to HCC and T cell leukemia, respectively, in an infected individual (335,336). Notably, oncogenic proteins of these viruses exploit or target one or multiple TRAF proteins for their signal transduction. These include the EBV-encoded oncoprotein LMP1, v-FLIP of KSHV, pUL48 of HCMV, E2 and E6 of HPV, Core protein of HCV and Tax protein of HTLV-1. In particular, consistent with the high frequency of TRAF3 deletions and mutations in HPV+ HNSCC, overexpression of TRAF3 inhibits the growth, migration and chemoresistance of HPV+ HNSCC by decreasing HPV E6 oncoprotein and increasing p53 and RB tumor suppressors (328). We briefly summarize the TRAF-dependent signaling mechanisms of pathogen-encoded proteins that contribute to human carcinogenesis as detailed in Table 3.

INDIRECT MECHANISMS OF TRAFS IN HUMAN CANCERS
Although beyond the scope of this review, we would like to point out that as critical regulators of adaptive immunity, innate immunity, and inflammation (1-4), TRAF proteins may indirectly contribute to the development, progression, and metastasis of various cancers by affecting tumor surveillance, tumor immunity, chronic inflammation and the tumor microenvironment. For example, disorders of innate antibacterial response are of fundamental importance in the development of gastrointestinal cancers, including pancreatic cancer, and increased expression of TRAF6, TLR4, and NOD1 are detected in peripheral blood leukocytes of pancreatic cancer patients (337). Specific deletion of TRAF3 from myeloid cells leads to development of B lymphomas and liver cancer in mice (56,174). Similarly, lymphocytespecific TRAF3 transgenic mice develop autoimmunity, inflammation and cancers (such as squamous cell carcinomas of the tongue, salivary gland tumors, and hepatoma) (55). TRAF2 regulates inflammatory cytokine production in tumorassociated macrophages, which facilitates tumor growth (46). TRAF4 promotes lung cancer aggressiveness by modulating the tumor microenvironment in normal fibroblasts via the TRAF4-NOX2/NOX4/p47-phox-ROS pathway (338). Importantly, TRAFs are also recognized as potential targets or modulators of cancer immunotherapy. For example, the immune adjuvants dsRNA such as Sendai Virus, poly-I:C, and rintatolimod all activate the TLR3-TRAF3-IRF3 axis to promote CD8 cytotoxic T lymphocytes chemotaxis to the tumor microenvironment in cancer immunotherapy (339). Anti-GITR immunotherapyinduced tumor-specific Th9 cells, which are highly effective in eradicating advanced tumors in vivo, display a unique hyperproliferative feature driven by the Pu.1-TRAF6-NF-κB axis (340,341). Furthermore, TRAF3 and TRAF6 are crucial for osteoclast differentiation, and therefore can regulate bone metastasis of various cancers (4,342). We shall witness rapid advancement in these exciting areas in the coming years.

CONCLUSIONS
In this article, we have analyzed the current evidence of genetic alterations of the TRAF family in human cancers. The results revealed that genetic alterations of all seven TRAF genes are present in various human cancers and that recurrent mutations of each TRAF gene have been detected in cancer patients. In particular, loss-of-function genetic alterations of TRAF2 and TRAF3 are frequently detected in B cell malignancies, and the rates of missense mutations of TRAF7 are overwhelmingly high in adenomatoid tumors, secretory meningiomas and perineuriomas. Gain-of-function alterations (gene amplification and overexpression) are common for TRAF1, TRAF4, TRAF5, and TRAF6 in human cancers, and are also identified for TRAF2 in epithelial cancers. Corroborating human evidence, direct causal roles of TRAF genetic alterations (except TRAF7) in tumorigenesis have been demonstrated in vivo with genetically engineered mouse models that have each TRAF gene deleted or overexpressed in specific cell types. Importantly, however, the functional significance of most TRAF point mutations identified in human cancers remains to be assessed in future studies. A number of interesting TRAF-dependent oncogenic and tumor suppressive pathways have been elucidated from both in vivo and in vitro studies, although current understanding is still far from complete and further investigation is required. The significance of TRAFs in cancer pathogenesis is reinforced by the evidence that TRAF proteins also participate in pathogen-induced carcinogenesis, including bacteria and viruses. Furthermore, emerging evidence indicates that TRAF proteins can indirectly contribute to tumorigenesis and metastasis by affecting tumor immunity, chronic inflammation, bone resorption, and the tumor microenvironment. In conclusion, the information presented in this article provides a rationale for the development of novel immunotherapies and other strategies to manipulate TRAF proteins or TRAF-dependent signaling pathways in human cancers by precision medicine, which represents the next primary challenge in the field.

AUTHOR CONTRIBUTIONS
PX and SZ have taken the leading roles in designing and writing this manuscript, and all co-authors (JJ, SG, AL, HS, and JF) have also made significant contributions to writing this manuscript.