NFκB signaling in T cell memory

Memory T cells play an essential role in protecting against infectious diseases and cancer and contribute to autoimmunity and transplant rejection. Understanding how they are generated and maintained in the context of infection or vaccination holds promise to improve current immune-based therapies. At the beginning of any immune response, naïve T cells are activated and differentiate into cells with effector function capabilities. In the context of infection, most of these cells die once the pathogenic antigen has been cleared. Only a few of them persist and differentiate into memory T cells. These memory T cells are essential to host immunity because they are long-lived and can perform effector functions immediately upon re-infection. How a cell becomes a memory T cell and continues being one for months and even years past the initial infection is still not fully understood. Recent reviews have thoroughly discussed the transcriptional, epigenomic, and metabolic mechanisms that govern T cell memory differentiation. Yet much less is known of how signaling pathways that are common circuitries of multiple environmental signals regulate T cell outcome and, precisely, T cell memory. The function of the NFκB signaling system is perhaps best understood in innate cells. Recent findings suggest that NFκB signaling plays an essential and unique role in generating and maintaining CD8 T cell memory. This review aims to summarize these findings and discuss the remaining questions in the field.


Introduction
The transcription factor Nuclear Factor kappa B (NFkB) regulates many aspects of the innate and adaptive immune system. Innate cells signal through NFkB to drive inflammation, while in lymphocytes, NFkB signaling supports their activation, differentiation, survival, and effector function. Perhaps due to NFkB's varied and multiple effects, it has been challenging to dissect how and when NFkB signaling regulates each of these processes in T cells. Particularly in the context of the immune responses that accompany diseases such as autoimmunity, cancer, or infection. This review will focus on recent findings indicating that NFkB is critical for generating and maintaining T cell memory. The role of NFkB signaling in inflammation and lymphocyte activation has been recently reviewed in (1)(2)(3). While NFkB signals can regulate innate immune cells and immune responses and indirectly affect the differentiation of T cells to memory, here we will focus on the T cellintrinsic mechanisms by which NFkB supports T cell memory and the environmental cues that drive them.

The NFkB signaling pathway
NFkB signaling is classically described as canonical and noncanonical. In T cells, recognition of antigen (Peptide-MHC) by the T cell receptor (TCR) and cluster of differentiation (CD) 28 (CD28) co-stimulation is required to activate the canonical NFkB signaling pathway. TCR engagement leads to the phosphorylation of the CD3 immunoreceptor tyrosine-based activation motifs (ITAMs) and recruitment of the zeta chain of T cell receptor associated kinase  to the plasma membrane, where it can, in turn, phosphorylate the adaptor molecules linker of activation for T cell (LAT) and SH2 domain-containing leukocyte phosphoprotein of 76kDa (SLP-76) enabling the assembly of the LAT/SLP-76 signalosome. Phosphorylated LAT binds to interleukin-2 inducible T-cell kinase (Itk) and phospholipase C g1 (PLCg1) and allows for Itk to phosphorylate and activate PLCg1, which can then hydrolyze Phosphatidylinositol 4,5-bisphosphate (PIP2) in the membrane to generate two second messengers: inositol triphosphate (IP3), which binds to its Endoplasmic Reticulum receptor and triggers calcium signaling; and diacylglycerol (DAG) which is critical for the activation of protein kinase C q (PKCq) (3) (Figure 1).
PKCq is a crucial enzyme for TCR-dependent canonical NFkB signaling. In resting conditions, PKCq remains in the cytosol. However, upon TCR stimulation, SLP-76 phosphorylation allows for the recruitment of the guanine nucleotide exchange factor Vav, which helps to activate Rac GTPase and reorganize the cytoskeleton in a process that brings PKCq from the cytosol to the membrane where, by the coordinated action of DAG, the GCK-like kinase (GLK) and 3-phosphoinositide-dependent protein kinase-1 (PDK-1), PKCq is activated and connected to downstream signaling networks (CBM complex, explained below) (4)(5)(6)(7). In the membrane, PKCq is located in specific microclusters or lipid rafts where TCR and CD28 receptors are enriched. Of note, both antigenic and B7/CD28 signals are required for PKCq activation, making PKCq the first molecular hub where antigenic and costimulatory signals integrate to induce NFkB (8) (Figure 1).
The next link in the NFkB signaling cascade is CARDcontaining MAGUK protein 1 (Carma1). Carma1 is a membrane scaffold protein recruited to lipid rafts after TCR stimulation. PKCq, phosphorylates Carma1, which undergoes a conformational change and binds to another two molecules, B-cell lymphoma/leukemia 10 protein (Bcl10) and mucosa-associated lymphoid tissue lymphoma translocation protein 1 (Malt1), to form the CBM complex. The CBM complex acts next as a core where different adaptors, ubiquitin ligases, and enzymes ultimately activate the IkB kinase (IKK) complex (9,10). In brief, activated Carma1 recruits the linear ubiquitin chain assembly complex (LUBAC) complex leading to the poly ubiquitination of Bcl10 and Malt1 via the action of the TNF receptor associated factor (TRAF) 6 ligase (TRAF6). This is followed by the activation of the Transforming growth factor-b (TGF-b)-activated kinase 1 (Tak1) and the recruitment of the IKK subunits (IKKa, IKKb and IKKg or NEMO). The proximity of Tak1 to the IKK subunits allows Tak1 to phosphorylate IKKb. Phosphorylated IKKb, phosphorylates the protein nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, alpha (IkBa), which is bound to the transcription factor NFkB in the cytosol. Finally, upon phosphorylation, IkBa is targeted for degradation in the proteasome, leaving NFkB free to translocate to the nuclei, where it regulates different aspects of gene expression [reviewed in (11) (discuss later) ( Figure 1).
Non-canonical NFkB signaling is classically associated with tumor necrosis (TNF) receptor superfamily (TNFRSF) signaling. In resting conditions, NF-kB-inducing kinase (NIK) is continuously targeted for degradation upon the action of cellular inhibitor of apoptosis (cIAP) and TRAF2,3 ligase and adaptor activities. However, upon TNFRSF stimulation, cIAP mediates the K48 polyubiquitination and degradation of TRAF3. This allows for the accumulation of NIK and its activation, resulting in p100 processing, p52 production and binding to RelB (12) (Figure 1). Importantly, NFkB comprises various homo and heterodimers whose composition differs between the canonical and non-canonical NFkB pathways. The canonical NFkB pathway involves p50/p65 (RelA) and p50/c-Rel heterodimers, where p50(NFkB1) results from the constitutive degradation of p105 (13). In contrast to this, noncanonical NFkB signaling leads to the nuclear entry of the p52/RelB heterodimer, where p52 results from p100 processing. Another fundamental difference between canonical and non-canonical NFkB signaling is the composition of the IKK complex. While canonical signaling requires IKKb, non-canonical NFkB signaling does not. Instead, the kinases NIK and IKKa are necessary to process p100 (the equivalent of IkBa) into p52 (12) (Figure 2).
It is widely accepted that antigen receptors do not trigger the noncanonical NFkB pathway (p100 processing and p52-RelB nuclear translocation). However, recent reports have challenged this view and shown that antigenic/TCR stimulation can also lead to non-canonical NFkB signaling (14)(15)(16)(17). It is unclear whether the link between TCR and non-canonical NFkB signaling occurs directly or indirectly through TCR-dependent induced expression of TNFRSF members CD27, OX40, or 4-1BB (18). Contrasting this idea, Yu et al. showed that anti-CD3 stimulated T cells were able to induce non canonical NFkB members p100 and its mature form p52. Furthermore, they used NIK-deficient T cells and revealed that induction of p52 occurred much earlier than the time required for the induction of TNFR. These cells exhibited an activated phenotype that resembled effector and memory T cells (15). Non canonical NFkB signaling can also influence canonical NFkB signaling at least in two ways. On one hand by increasing the expression of RelA/RelA homodimers that are then activated by canonical signaling and can contribute to inflammation in the gut (19). On the other, by generating p52/RelA heterodimers that can reinforce RelA canonical NFkB activity (20). Both of these processes have been described in epithelial cells but a role in T cells remain yet to be explored.
Another fundamental difference between canonical and noncanonical NFkB signaling is the kinetics of their activation. While the canonical pathway leads to rapid but transient activation of NFkB (21)(22)(23), the activation of the non canonical pathway is slow and persistent (12). Finally, it is important to consider that crosstalk between both canonical and non-canonical pathways, not only in term of positive but also negative feedback loops,has been described (18) reviewed in (24,25)), making it challenging to understand how this important signaling pathway regulates the many responses and differentiation processes of a T lymphocyte.

NFkB, T cell memory and T cell memory responses
During an immune response to pathogens, tumors, and self or transplanted organs, both naïve CD4 and CD8 T cells use their T cell receptors (TCRs) to recognize antigenic peptides associated NFkB family members, homo and heterodimers and their roles in T cells. Second column shows the five NFkB subunits, p50, RelA, c-Rel, p52 and RelB that associate with each other to form fifteen potential dimers. Of these, three heterodimers (p50/RelA; p50/c-Rel and p52/RelB) are well known for their roles in T cell function and differentiation, while the role of the remaining five homodimers and seven heterodimers is unknown in T cells (indicated by? in the table). A role in other cells is indicated in the table. The table also shows whether any of the dimers is preferentially induced by canonical or non-canonical NFkB signaling (column 4) or whether it is the result of processing of precursor proteins p100 and p105 (column 1).
with major histocompatibility complex (MHC) molecules (p-MHC) on the surface of dendritic cells (DCs) in secondary lymphoid organs such as lymph nodes or spleen. Antigen recognition, together with costimulatory and pro-inflammatory signals, provide the basic information that is required for a T cell to differentiate into effector and memory T cells. Effector and memory CD8 T cells are directly in charge of eliminating the cells that are the source of the antigen (i.e., infected or cancer cells) while CD4 effector and memory T cells indirectly contribute to the clearance of pathogens or tissues that are source of the antigen (via the activation or differentiation of other immune cells) (26, 27) ( Figure 3). Both effector and memory CD4 and CD8 T cells do so through a series of changes in their epigenetic and transcriptional landscape that enable them to exert effector functions such as cytotoxicity or cytokine secretion. Yet a fundamental feature of memory T cells is their capacity to survive for long periods of time in the host and their ability to sustain their effector capabilities. The capacity to survive is a rare and unique characteristic in the responding T cell population since only 1-10% of the T cells recruited into the immune response can persist during the whole immune response and become memory (28) (Figure 3). Indeed, how memory T cells are generated and maintained it is still not fully understood. Importantly, memory T cells defend us against infection and cancer but also contribute to autoimmunity and transplant rejection. Therefore, thoroughly defining the mechanisms that regulate their generation and maintenance has important implications for immune therapies, including vaccines and immunotherapies targeting cancer, autoimmunity and transplantation.
The tremendous progress in the field in the last 20 years has given us a clearer idea of the transcriptional and epigenetic mechanisms that drive T cell memory. Since more is known about these processes in CD8 than CD4 T cells, from here on we will focus on CD8 T cells.
CD8 T cell memory is not monochromatic. Rather, CD8 memory is very diverse, comprising cells that stay in tissue (tissue-resident or T RM ) and others that remain in the circulation, either in lymphoid organs (central memory or T CM ) or in the vasculature as peripheral effector memory T cells (T EM ) (29). This diversity promises to be even broader, with more specialized phenotypes being identified yearly (30, 31). Interestingly, each T cell memory subset's survival and effector capabilities also appear to be distinct. For example, central memory T cells persist the longest and proliferate vigorously upon recall but are less efficient at effector responses upon recall than effector memory. By contrast, tissueresident memory T cells in specific tissues, such as the lung, are short-lived but can quickly mount a vigorous response within the tissue (29).
In the search for biochemical mechanisms that explain how memory T cells arise, a part of the field has focused on how antigenic and inflammatory cues are transduced to signal the maturation toward memory. In this line, studies in the early 2000s determined mammalian target of rapamycin (mTOR) signaling was a major pathway used by inflammatory signals such as Interleukin-12 (IL-12) to limit the ability of effector T cells to progress to memory by increasing the ratio of transcription factors T-bet to Eomes (32). By contrast, Wnt/Phosphoinositide 3-kinase (PI3K) signaling turned out to be essential to favor the generation of memory T cells, in part by increasing the levels of tumor cell factor 1 (TCF-1) and Eomes and decreasing the T-bet/Eomes ratio (33)(34)(35). Another important finding was the realization that metabolic changes are also necessary for T cells to become memory and that these changes were modulated by the intensity of antigenic and inflammatory signals (36-39).
NFkB is a pleiotropic signaling pathway with important roles in controlling inflammation, cell activation, and cell survival. Studies by Schmidt-Supprian, Rajewsky and Pasparakis determined the crucial relevance of NFkB signaling to maintain the survival of naïve T cell (40). Multiple studies also established that NFkB signaling is required for early T cell activation, cell division, cytokine production, and cytotoxicity (2). Using animal models whose T cells were devoid of several signaling intermediates in the NFkB signaling cascade, studies found that this pathway is especially relevant for the differentiation and/or effector function T cell memory differentiation and environmental triggers of NFkB signaling Naïve T cells recognize antigen through their TCR. TCR and CD28 signaling induce the NFkB pathway early in the response. This pathway is necessary for the cell to proliferate and acquire effector function. These signals together with other costimulatory signals from TNF receptors superfamily also signal to NFkB to sustain proliferation and effector function. In addition, inflammatory signals from cytokines and TLR could also contribute to shape the T cell intrinsic dynamics of NFkB signaling. Once effector T cells are generated they eliminate the cells source of antigen and inflammation and most of T cells (SLECs or short lived effectors) die by apoptosis. The MPECs or memory precursors continue their differentiation to memory. For this, an interplay between epigenetic and transcriptional changes allow for silencing of specific gene and poising others for quick effector gene reactivation. In addition, positive NFkB feedback loops allow for establishing memory fitness. The timing of these changes during T cell differentiation in relationship with a given T cell immune response are depicted in the figure through a graph of the kinetics of a T cell immune response to a hypothetical pathogen. Days of the immune response post infection are represented on the X axis. Y axis shows number of antigen specific T cells responding to the pathogen. Yellow blocks indicate the triggers of NFkB input over the course of the immune response.
of CD4 T helper (Th) Th1, Th2, Th17, Tfh and regulatory T (Treg) cell subsets (1,(41)(42)(43). The role of NFkB in T cell memory has been less explored, perhaps because NFkB is critical for most of the early aspects of T cell activation, proliferation, and effector differentiation, making it challenging to dissect its unique roles in T cell memory. One study, however, reported that CD8 T cells uniquely impaired in triggering TCR-dependent NFkB signals failed to differentiate into memory T cells in response to infection. The memory defect correlated with impaired RelA and c-Rel transcriptional activity. Yet, these NFkB-impaired CD8 T cells did not exhibit any defects in proliferation or effector function.This work indicated that the antigenic signals could control CD8 T cell memory through NFkB (44). Follow-up studies determined that inhibiting T cell-intrinsic NFkB signaling exclusively after the peak of the immune response to Listeria monocytogenes infection (once T cells have acquired effector function) dramatically hindered the ability of T cells to express the memory-associated transcription factor Eomes (45, 46). As a consequence, CD8 T cells failed to transition to memory. This was true for memory T cells in circulation, with inhibition of NFkB affecting mainly memory T cells of a central memory phenotype. Interestingly, the study also suggested that once memory T cells are formed, they also depend on NFkB signaling for their maintenance, as pharmacological inhibition of NFkB decreased B-cell lymphoma 2 (Bcl2) expression and led to a loss in memory fitness (46).
It is still poorly understood whether NFkB regulates CD4 T cell memory in the same manner or if its requirement will differ depending on the specific T-helper subset. Likewise, the potential role of NFkB in the differentiation of resident memory T cells has remained elusive. Using inducible NFkB models, we have recently described that lung tissue-resident memory T cells need a continuous input of NFkB signals for their maintenance. Inducible activation of NFkB signals in influenza-specific lung CD8 T RM cells boosted their numbers, presumably by increasing CD122 or IL15Receptor B chain and Bcl2 levels. This also appears to be true for circulatory T CM and T EM cells. Strikingly, we found that the role of the NFkB signal in generating lung CD8 T RM (before the T RM pool has been established) is different and unique. Inducible activation of NFkB signals in CD8 T cells transitioning to T RM led them to succumb. In contrast to this, T cell-intrinsic inhibition of NFkB signaling resulted in a considerable increase in the generation of influenza-specific CD8 T RM in the lung upon influenza infection. Remarkably, the same manipulation of NFkB signaling had the opposite effect on the generation of T CM and no effect on T EM (47). These results are surprising and show that NFkB signaling is a regulator of T cell memory subset diversity. Most importantly, these studies offer new avenues to boost T RM in the lung. This is especially relevant in the case of respiratory infections, and its vaccines as CD8 T RM are known to be short-lived and are indispensable for heterosubtypic protection (48).
While it may take days for naïve T cells to generate an immune response, memory T cells can start secreting cytokines and eliciting their effector function in a matter of hours. Studies by the Farber's group have shown that inhibition of NFkB activity prevented early memory T cell signaling and TCR-mediated effector function, suggesting that NFkB is required for early recall responses (49).
Speed is not the only component of a healthy T cell memory response; the location of these cells also plays a role in the protection they provide. Another study showed that mice receiving lung memory T cells exhibited a rapid and enhanced viral clearance compared to those receiving spleen memory cells or naïve mice (50). Thus, maintaining T RM in tissue and their functions is crucial to provide long-term immunity, although whether the latter depends on NFkB signals remains to be determined. Next, we will discuss how different environmental cues required for the generation of T cell memory may modulate the levels of T cell-intrinsic NFkB signaling and their fate.
4 Extracellular signals that trigger NFkB signaling and can shape T cell fate Naïve T cells circulate between the lymph and blood, passing by secondary lymphoid organs (lymph nodes and spleen). They use their TCR to scan other cells' peptide-MHC molecules and detect the presence of foreign antigens they have not been tolerized against. This detection is based on interactions of TCR with peptide-MHC and triggers different signaling pathways (including NFkB) that are critical to 'prime a T cell' for activation, proliferation, and secretion of IL-2. Once a T cell is activated, it leaves the secondary lymphoid organ where it first encountered antigen and travels through the circulatory system to tissues where it can finally exert its true effector functions (Figure 3). Already at priming and later, when in tissue, a T cell can continue receiving antigenic/TCR signals together with other signals, such as costimulation via CD28 or TNFR superfamily members (such as CD40/CD40L, OX40, 4-1BB, Herpes Virus Entry Mediator (HVEM), or CD27). In addition, some inflammatory cytokines secreted by other cells in tissue can bind to receptors on T cells and signal to NFkB. Finally, T cells can also express and signal through TLRs (51). Some of these molecules are known for activating the canonical NFkB signaling pathway. However, others use the noncanonical NFkB signaling pathway (51, 52) ( Figure 1). Yet how a T cell integrates all these NFkB signal inputs across time or depending on the T cell differentiation status is unclear. Likewise, how a T cell interprets differences in dose, duration, and amplification of these NFkB inputs is a complex issue that remains largely unknown. It is also unclear whether previous stimulatory events are preserved in a T cell's molecular memory and pre-conditioning its subsequent response to the same stimuli as it has been recently reported for fibroblasts (53). Next, we will discuss in more detail the main environmental signals that can support T cell-intrinsic NFkB signaling as a T cell matures to memory.

Tumor necrosis factor receptor superfamily
Members of the TNFR superfamily signal through NFkB mainly through the canonical pathway (54). CD30, Lymphotoxin ab (LTab), B-and T-lymphocyte attenuator (BTLA), CD160, OX40, 4-1BB, CD27, Glucocorticoid-induced TNFR-related protein (GITR), death receptor 3 (DR3), and HVEM expression are induced on the surface of T cells once they have become activated (54,55) (56), and their expression can be maintained for hours or even days [review in (54)]. When they interact with their ligands (which are expressed on DCs, B cells, or macrophages (57)(58)(59) or non-immune cells in response to inflammation), these receptors can positively regulate CD4 and CD8 T cell responses and affect the outcome of disease (60). The consequences of TNFR-ligand interaction are bidirectional, as cells expressing the ligand also become activated. Given that T cells can express both receptors and ligands, this suggests that T cells within a population can potentially influence each other's responses (61).
In general, TNFRSF members regulate T cell responses in two ways. On one side, they support the production of cytokines directly on the T cell or indirectly by inducing the secretion of proinflammatory cytokines by antigen presenting cells (APCs). On the other side and most relevant to this review, TNFRSF members regulate the frequency of effector and/or memory T cells generated after antigen priming both in primary and recall immune responses. They do so via NFkB by regulating both the proliferative and survival capacities of the T cell [review in (54)]. This has been described for 41BB, CD27, TNFSF14 or LIGHT and GITR (62)(63)(64)(65)(66)(67), which support the expression of anti-apoptotic factors B-cell lymphoma-extra large (Bcl-xL), Bcl2 and BFL-1 (68). Most recently, 4-1BB has also been shown to modulate T cell mitochondrial metabolism to promote survival and persistence of memory T cells (69, 70).

Cytokines that signal through NFkB and their effects on T cell memory
Inflammatory cytokines, TNF, interleukin 6 and 1 b (IL-6, and IL-1b) levels increase over the course of infection in specific tissues where pathogen-specific T cells are present. T cells express receptors that can bind to these cytokines and, in this manner, increase the input of signals to NFkB. As a pro-inflammatory cytokine, TNF is known for inducing apoptosis in highly activated effector T cells. T cells, themselves, can express membrane-bound TNF, although this form of TNF needs to be cleaved to be most effective (71). T cells also express TNF receptor 2 (TNFR2) and can interact with TNF early upon antigen recognition. In this way, TNF acts as a costimulatory signal and can lower the threshold of TCR signaling to support proliferation, survival, and cytokine production via NFkB (72,73). TNF can also induce apoptosis on highly activated T cells, such as effectors, during and after the peak of an immune response (the contraction phase- Figure 2). For this, TNF needs to interact with TNF receptor 1 (TNFR1) on the surface of T cells. Curiously, apoptosis of T cells during the contraction phase of the response to Lymphocytic choriomeningitis virus (LCMV) is impaired when T cells lacked both TNF receptors, TNFR1 and TNFR2 (74). By contrast, selective deletion of TNFR2 appears to enhance the persistence of CD8 T cells maturing to memory (75). These findings suggest a delicate balance between TNFR1 and TNFR2 signals regulates T cell memory. (Figure 1).

IL-1b triggers the canonical NFkB pathway and can also modulate T cell responses. T cells can indeed express IL-1 receptor (IL-1R), and when they bind IL-1, this interaction can enhance T cell proliferation and effector responses of both naïve and memory T cells (76, 77)
IL-6 has also been shown to induce NFkB signaling in intestinal epithelia (78), although whether IL-6 signaling in lymphocytes also induces NFkB remains to be defined. Most recently, the Medhitovz group has shown that IL-6 and IL-1 cooperate to control the inhibitory effects of regulatory T cells on CD4 T cells and that IL-6 is critical for the generation of memory CD4 T cells. Yet whether this is dependent on NFkB signaling, relates to T cell survival, or has a similar effect on memory CD8 T cells is still unclear (79).
In summary, cytokines are often considered major Janus kinase (JAK)-signal transducer and activator of transcription (STAT) pathway (JAK/STAT) signaling users to shape CD4 T helper differentiation. However, a few of the inflammatory cytokines (IL-6, IL-1, and TNF) present in abundance in disease scenarios such as infection, chronic inflammation or therapies such as chimeric antigen receptor (CAR) immunotherapies (80) can also limit the generation of long-term protective T cell memory and may do so through NFkB signaling. The ability to trigger NFkB signals makes them unique as they could contribute to shaping the overall NFkB system of the cell and alter its ultimate fate.

Toll like receptors
TLRs are pattern recognition receptors that recognize microbial molecules characteristic of specific pathogens. They are expressed in human and mouse sentinel cells such as DCs and macrophages. TLRs are comprised of 13 different receptors (TLR1-13), some of which are not shared by humans and mice, and some are expressed upon TCR stimulation (51). TLRs classically trigger canonical NFkB signaling via Tak1 (52, 81). Upon ligand binding, TLRs dimerize, and their TIR domains are recognized by receptor-proximal membrane molecules The Toll-interleukin-1 Receptor (TIR) domain-containing adaptor protein (TIRAP) or TRIF-related adaptor molecule (TRAM), which enable the assembly of two different scaffold supramolecular organizing centers (SMOCs). TIRAP seeds the Myddsome while TRAM seeds the Triffosome. The Myddsome includes Myeloid differentiation primary response 88 (MyD88) and Interleukin-1 receptor-associated kinase (IRAK) kinases, which, once activated, recruit the ubiquitin ligase TRAF6. TRAF6 mediates the activation of Tak1, thereby inducing NFkB signaling. TRAF6 also mediates the activation of TANK-binding kinase 1 (TBK1), which can also feedback on NFkB signaling and activate glycolysis via protein kinase B or Akt. (reviewed in (82) (Figure 1). TLR2, TLR5, and TLR9 work as costimulatory signals on effector T cells, but only when antigen recognition is in place. As co-stimulatory molecules, they support proliferation and effector function (51). Instead, T cell-intrinsic TLR2 signaling has recently been shown to regulate CD8 T cell survival and CD8 T cell memory. In the context of viral infections, TLR2-Myd88 deficient CD8 T cells poorly differentiate into memory T cells due to an inability to survive during the contraction phase of a vaccinia virus infection (83). The authors suggested that this depended on PI3K activity although a contribution of NFkB signaling still remains to be explored (83). Another study showed that CD8 T cells stimulated with low-affinity antigen and TLR2 ligands more efficiently generated memory phenotype CD8 T cells than without TLR costimulation. These memory-type CD8 T cells could respond and survive in vivo (84). While less is known for other TLRs, it is tempting to hypothesize that TLR can have an important role in modulating T cell memory either directly depending on the signal input the T cell receives or indirectly via the regulation of the function of other immune cells.

Intracellular signals that feed into NFkB signaling and can shape T cell fate
NFkB signaling can also be induced in a T cell without involving extracellular receptors. An obvious example is intracellular TLR3,7, and 9, which are expressed on endosomal membranes and signal to NFkB (52). Yet aside from TLR, other innate sensors able to signal to NFkB have been recently shown to be expressed inside T cells. The one that has drawn most attention recently is Stimulator of interferon genes(STING). STING is an adaptor molecule that binds to cyclic di-nucleotides generated by intracellular pathogens or cyclic guanosine monophosphate (cGMP) generated as a consequence of the activation of the Cyclic GMP-AMP synthase (cGAS) once it has recognized cytosolic DNA. STING, together with TLRs, supports the activation of NFkB and interferon b (IFNb). For STING signaling, the recruitment of TBK1 to STING allows both the activation of the IKK complex and induction of NFkB and the activation of interferon regulatory factor 3 (IRF-3) (85). We have recently described that T cell-intrinsic STING signaling can tightly regulate CD8 T cell memory (86). During the course of a T cell immune response, we found that excessive STING signaling hindered the generation of CD8 T cell memory but this was reverted when T cells lacked STING. Our studies also found that TCR signal strength regulated the level of induction of STING signaling within effector CD8 T cells when they were exposed to STING agonists. Strong antigenic signals supported the maximum level of STING signaling when T cells were exposed to STING agonistic stimulation. In turn, high levels of STING signaling upregulated the expression of the integrated stress response (ISR) molecule C/EBP homologous protein (CHOP), which caused Bcl-2 Interacting Mediator of cell death (Bim)mediated apoptosis in T cells transitioning to memory (86). We believe these findings are especially important in the context of current STING agonistic therapies developed for cancer treatment. They suggest tumor-specific T cells may be lost if STING agonist levels are not properly dosed. The data also has important implications for the treatment of infections that cause persistent activation of STING or inflammatory diseases where STING signaling is aberrantly overactivated (87)(88)(89)(90). In these cases, treatments that attenuate STING signaling might be especially critical to regulate the deleterious effects of inflammation and to preserve the generation of T cell protective immunity (91).
It is known that NFkB signaling can be subjected to negative feedback loops via the regulated expression of inhibitors IkBa, IkBdand IkBe (92, 93). These negative feedback loops establish a different signaling modality from the typical transient/strong or delayed/sustained NFkB activation. They introduce the idea of oscillation which can be interpreted by the cell as a completely different input using the same signaling pathway (94). Little is known, however, of positive feedback loops that can perpetuate signaling without external input. It has been proposed that positive feedback loops in a signaling network (or bi-stability) can create a memory of a transient differentiating stimulus long after the stimulus is removed by establishing self-sustaining patterns of gene or protein expression (95), thereby providing a framework of how cell differentiation can be achieved and lineage commitment can be maintained (96). Most recently, such a positive feedback loop has been shown for NFkB signaling in the context of T cell memory (46). One study described a novel mechanism of how commitment to CD8 T cell memory could be achieved by perpetuating NFkB signals via a feedback loop long after the original TCR-dependent NFkB signaling had been removed. In their study, the authors reported how the TCR-dependent NFkB signal delivered early in the immune response led to the expression of a kinase, Pim-1 protooncogene, serine/threonine kinase (Pim1K), that was essential to maintain the induction of NFkB signals late in the response while T cells were differentiating to memory. Impairing the activity of any of the members of the feedback loop resulted in a loss of the memory T cells generated upon infection, indicating that this NFkB positive feedback loop is critical to T cell memory fitness (46).
Taken altogether, it is becoming clear that during the course of an immune response, T cells are subjected to a myriad of extracellular and intracellular inputs that feed into the NFkB signaling system. These inputs might be wired in different ways to generate distinct responses and a memory of the original stimuli that is perpetuated to enable T cell differentiation and final fate commitment (53).

NFkB signaling and T cell memory transcriptional and epigenetic programming
The NFkB signaling pathway leads to the induction of the transcription factor NFkB. As mentioned before, NFkB is a dimer that can be composed of the combination of two of five different subunits, p65(RelA), p50 (NFkB1), c-Rel, p52 (NFkB2), and RelB ( Figure 2). All subunits share a rel homology binding domain (RHD) necessary for DNA binding and binding to other subunits. In addition, each subunit contains transactivation domains responsible for transcriptional activity[review in (42)]. Much is known about the signaling intermediates that lead to NFkB nuclear translocation and the external stimuli that support this process, but how NFkB regulates endogenous target genes has remained elusive.
Similarly, why and how NFkB leads to different gene expression patterns depending on the cell type has yet to be elucidated. Two main features (or outcomes here) define a memory T cell. On one side, the ability to quickly re-activate effector functions. On the other side, the unique ability to have a long life. Resolving how T cell memory is regulated involves understanding whether the same mechanisms control these two features or not. In the case of T cell memory, it is important to consider that the genetic program a memory T cell inherits comes from previously activated effector T cells. Add to this the idea that T cell memory differentiation continues long after antigen is cleared, and the possibility to integrate environmental cues from tissue, and it is easy to perceive the level of complexity in memory programming.
A way that may help to tease this apart is to consider that the capacity to be long-lived and the capacity for quick induction of effector function can be differentially regulated by NFkB. If we consider the transcription factors that are now well known for their ability to regulate T cell memory differentiation[recently reviewed in (97)], one can observe that some have a more significant role in memory survival than in effector function differentiation. For example, in the T-bet/Eomes tandem, increasing Eomes levels over T-bet is necessary to support Bcl-2-dependent survival (98). In the B lymphocyte-induced maturation protein-1(Blimp-1)/(Bcell lymphoma 6)Bcl-6 tandem, however, Blimp-1 is more involved in effector function and Bcl-6 in memory fitness. NFkB regulates both Eomes and Blimp-1. Yet, TCR-dependent NFkB signaling deficient CD8 T cells, are deficient in Eomes but not in T-bet or Blimp1 expression. They can differentiate into effectors but fail to progress to memory (46), indicating that TCR-dependent NFkB signals uniquely utilize NFkB signaling to program the longevity of memory T cells. Interestingly, NFkB directly regulates the expression of Eomes (99) but the NFkB -dependent regulation of Blimp-1 may follow a different mechanism. Thus, establishing a program of memory longevity might involve the transcriptional ability of p65 NFkB alone or with specific players such as Notch to regulate Eomes expression (99, 100) but the same mechanisms may not control Blimp-1.
When considering the different T cell memory subsets, one can only wonder whether the same NFkB mechanisms apply to all. Eomes and TCF-1 are master regulators of central memory T cells in the circulation. However, too much Eomes expression and too little T-bet expression prevents T cells from becoming tissueresident memory (97). Instead, the induction of the transcription factor Runx3 appears more crucial for T RM homeostasis (101). In this case, Eomes may be dispensable for T RM survival although NFkB signaling may still have a role in CD8 T RM by regulating in a different manner Runx3 expression. One possibility is via the Pim1K axis as Pim1K is required for Runx3 nuclear translocation (46,102).
Other mechanisms may be also in place to regulate other molecular players of T cell memory and explain how NFkB signaling leads to different T cell fates (Figure 4). Some T cell outcomes may be determined using the canonical versus noncanonical pathway and the induction of specific NFkB heterodimers ( Figure 2). Additionally, the accessibility of NFkB to specific promoters depends on the cell's open vs. closed chromatin status (103,104). The ability of NFkB to synchronize with other cofactors may also contribute to specific gene expression in T cells differentiating to memory (105). Finally, the dynamics of the NFkB response can vary depending on the frequency and type of extracellular signal input, which might be translated by the cell into different transcriptional responses or no response at all. For example, NFkB signals can be oscillatory as long as a stimulus can produce waves of NFkB nuclear entry under the NFkB-dependent expression of the IkBa feedback loop. Alternatively, NFkB oscillations may be dampened by the expression of alternative IkB inhibitors that are not susceptible to NFkB control. Oscillatory and dampened NFkB signal modes have been shown to regulate different genes depending on mRNA half-life and chromatin-regulated mechanisms (104,106,107). Although this has not been explored, it is possible that some of the extracellular cues a T cell samples in circulation or tissue can deliver distinct NFkB nuclear dynamics, thereby explaining why only certain NFkB -dependent genes will be expressed. Another possibility that could A B FIGURE 4 Mechanisms by which T cell intrinsic NFkB signaling could contribute to establish T cell memory outcome. explain NFkB -dependent T cell outcome diversity is the unique contribution of the two transactivation domains of the p65 NFkB subunit depending on posttranslational modifications as has been recently described for primary fibroblasts (108). In summary, NFkB is a mediator of a complex signaling circuit able to integrate multiple environmental cues over time and interpret them into specific T cell outcomes, including T cell memory ( Figure 4). However, a thorough understanding of how these mechanisms operate in T cells is still missing. Pioneering work using mathematical models and system approach strategies is revealing some of these mechanisms for innate cells (53,(109)(110)(111). These exciting studies may have drawn out the map to solve the unresolved questions of how NFkB regulates T cell outcome.

Alternative IKK/NFkB signaling triggered by environmental cues
While it is often assumed that the NFkB signaling cascade exclusively induces the transcriptional activation of NFkB, increasing evidence demonstrates that intermediates in this pathway can alternatively lead to the activation of other signaling pathways and the induction of transcription factors in an NFkBindependent manner. Already when PKCq function was described, it was evident that it could regulate both NFkB and MAPK activities. More recently, other signaling intermediates of the NFkB pathway, such as Carma1, have been shown to activate mTOR upon TCR stimulation (112). Malt1 and IKKs can activate MAPKs independent of NFkB induction (11, 113). In addition, environmental cues, such as cytokines and TLRs also use the NFkB machinery to trigger activation of other specific Mitogen-activated protein kinases (MAPK) signaling pathways (114)(115)(116).
There is little understanding of how alternative NFkB signaling modes regulate T cell responses and memory. Even less is known of whether the IKK/NFkB signaling pathway can be selectively directed to only support these alternative pathways and not NFkB transcriptional activation or vice versa, depending on the signal input from cytokines, antigens, or TLR ligands.
We propose a comprehensive integration of all of the 'flanks' of the NFkB signaling pathway is necessary to fully understand how this essential signaling pathway regulates T cell-mediated immunity and to include this knowledge into therapies that lead to better vaccines and T cell-based immunotherapies.

Conclusion
NFkB signaling is often associated with inflammation. Naïve T cells depend on the NFkB signal for their survival, while recently activated T cells proliferate and secrete IL-2 and interferon g (IFNg) in an NFkB -dependent manner. Most recently, it has been shown that NFkB signaling is a crucial signaling pathway for generating and maintaining CD8 memory T cells. NFkB signaling appears especially critical for antigenic signals to program memory fitness and maintenance even after infection has resolved. Yet there are still important gaps that need solving. Besides antigen, other inflammatory signals present during an immune response can signal to NFkB within a T cell. It will be important to define how and which aid in generating and maintaining T cell memory. We also do not clearly understand whether the level of NFkB signal a T cell could experience in the context of disease shapes T cell memory in different ways. Understanding this can be critical for the treatment and/or vaccination of patients either suffering from chronic inflammation or being subjected to anti-NFkB therapies t. Lastly, the field is still in its infancy in understanding how NFkB signaling can specifically regulate T cell outcomes. It is becoming increasingly clear that a more systemic approach based on mathematical models and foundational work in innate cells needs to be applied to T cells to better define how this signaling pathway can be exploited therapeutically to shape T cell-based protective immunity.

Author contributions
ET wrote and edited the manuscript as well as organized the review. MD edited and contributed to the discussion of the manuscript. DL contributed to the organization and discussion of the manuscript. All authors contributed to the article and approved the submitted version.