T‐cell exhaustion in HIV infection

Abstract The T‐cell response is central in the adaptive immune‐mediated elimination of pathogen‐infected and/or cancer cells. This activated T‐cell response can inflict an overwhelming degree of damage to the targeted cells, which in most instances leads to the control and elimination of foreign invaders. However, in conditions of chronic infection, persistent exposure of T cells to high levels of antigen results in a severe T‐cell dysfunctional state called exhaustion. T‐cell exhaustion leads to a suboptimal immune‐mediated control of multiple viral infections including the human immunodeficiency virus (HIV). In this review, we will discuss the role of T‐cell exhaustion in HIV disease progression, the long‐term defect of T‐cell function even in aviremic patients on antiretroviral therapy (ART), the role of exhaustion‐specific markers in maintaining a reservoir of latently infected cells, and exploiting these markers in HIV cure strategies.


| G ENER AL INTRODUC TI ON ON T-CELL E XHAUS TION
The T-cell activation paradigm proceeds in a highly organized process involving three signals consisting of antigen recognition, receptor costimulation, and a termination signal that are required for the tight regulation of a strong functional and proliferative response.
Signal one in T-cell activation represents the specific recognition through the T-cell receptor (TCR) of their cognate antigen presented by professional antigen-presenting cells (APCs) including dendritic cells, macrophages, and B cells.
Signal two is the costimulatory signal where receptors on the T cells bind to their counterpart ligands. CD28 is the primary costimulatory receptor for T cells that acts through interaction with its ligands, CD80 and CD86 expressed on APCs. Costimulation is essential for T-cell activation since signal one TCR/antigen recognition in the absence of signal two drives the cell toward an anergic and/or tolerogenic state. 1 Additional costimulatory receptors that can enhance T-cell activation include CD27, OX40 (CD134), ICOS (CD278), CD40L (CD154), CD226 that bind to CD70, OX40L (CD252), B7-H2 (CD275), CD40, and CD155/CD112, respectively, on APCs. 2 The concentration of costimulatory and ligand molecules can vary significantly such that either no positive signal is sent or a paired interaction provides a strong supporting signal two to the T cells. Soluble pro-inflammatory cytokines including IL-12 and type I interferons contribute to a fully activated T-cell response.
Signal three arises in the days following T-cell activation, the effector phase of the immune response, and eventually the elimination of the pathogen. Signal three is responsible for terminating the im-

| FUN C TIONAL PROFILE
In exhausted CD8 T cells, the increased expression of ICIs leads to a hierarchical loss of function that begins with reduced IL-2 secretion, cytokine polyfunctionality, and diminished proliferative potential. Functional defects progress to the loss in cytotoxic capacity and TNF-α secretion with highly exhausted cells having reduced ability to produce IFN-γ. [6][7][8] Although T-cell exhaustion was first identified in CD8 T cells, it is now accepted that CD4 T cells are also subject to exhaustion leading to reduced production of IL-2, IFN-γ, and TNF-α 9-11 along with reduced CD4 T-cell help. 12 In addition to persistent antigen and pathogen burden, the loss or reduction of CD4 T-cell help is an important factor that contributes to the initial establishment of exhaustion. [13][14][15] The precise nature of optimal CD4 T-cell help needed to antagonize exhaustion is unclear, but is likely to involve the production of IL-2 and IL-21 cytokines that support CD8 T-cell response both directly and indirectly through activation of APCs. 16 Aside from a blunted immune response against infected or cancerous cells, exhausted T cells have a poor or varied response to homeostatic cytokines including IL-7 and IL-15 responsible for the maintenance of memory T cells. In the case of complete clearance of a viral pathogen or in adoptive transfer studies to antigen-free mice, exhausted T cells are poorly maintained through self-renewal. [17][18][19] Instead, persistence of exhausted T cells occurs through continual antigen signals that promotes proliferation. 20 Exhausted T cells can persist in vivo for years but in the final stages of exhaustion with high antigen stimulation, there is a loss of the virus or tumor-specific cells through apoptosis. 7,21 Further characteristics of T-cell exhaustion include altered transcription factor expression, metabolic profile and epigenetic modifications that are distinct from other memory T-cell subsets.

| TR ANSCRIP TIONAL PROFILE
The transcriptional profile of exhausted CD4 and CD8 T cells are significantly different from all memory T-cell subsets. Although several transcriptional factors are correlated with exhausted T F I G U R E 1 T-cell immune checkpoint inhibitors and related ligands. TCR interaction with the antigenic peptide-MHC complex displayed by professional APCs delivers the primary signal for T-cell activation. The CD28 co-receptor and other costimulatory receptors enhance the T-cell stimulatory signal following interaction with their corresponding ligands. Immune checkpoint inhibitors including PD-1, CTLA-4, LAG3, TIGIT, TIM3, CD160, and 2B4 act to suppress T-cell signal following interaction with their related ligands expressed on APCs cells including NFAT, Batf, IRF-4, T-bet, Eomes, and Blimp-1, no master regulator of exhaustion has been identified. NFAT activation during chronic infection along with a corresponding low nuclear translocation of AP-1 induces strong transcriptional activation of ICIs including PD-1, LAG3, and Tim3. 22

| ME TABOLIC AB NORMALITIE S AND E XHAUS TION
The metabolic profile of T cells shifts from the use of the oxidative phosphorylation pathway in naive cells to glycolysis in the acute phase of an infection when activated effector cells have increased bioenergetics needs. Once the infection is cleared, memory cells revert to a quiescent state that uses oxidative phosphorylation and gains the additional metabolic ability for fatty acid oxidation. 37 With chronic antigen stimulation and the onset of exhaustion, the suppression of the glycolysis pathway ensues with reduced cellular glucose uptake and signs of a dysregulated mitochondrial function. 38 The utilization of endogenous fatty acids by exhausted cells may dictate the available energy reserves under conditions of ICIs engagement. 39 Metabolic pathways implicated in this defective state include transcriptions control through Foxo1 40 and PGC1α. 41  HIV, CMV, and EBV. 43 Indeed, the vast majority of this epigenetic program linked to exhaustion in mice (approximately 80%) was also successfully mapped to tetramer positive CD8 T cells of treatment naive chronically infected HIV donors. [44][45][46] As such, epigenetic modifications play a key role in maintaining T cells in an exhausted state.

| T-CELL E XHAUS TI ON IN HIV INFEC TI ON (C AUS E S OF E XHAUS TI ON , PHENOT YPI C MARK ER S OF E XHAUS TI ON , AND FUN C TIONAL E XHAUS TION)
T-cell exhaustion was first described in mouse models with chronic LCMV infection where antigen-specific CD8 T cells progressively lost their effector functions and developed a reduced capacity to kill virally infected cells. [47][48][49][50] Subsequent to these studies, it became clear that The progressive loss of CD8 T-cell function starts with an initial loss of proliferative capacity, cytotoxic potential, and a restricted IL-2 production. In more pronounced stages of exhaustion associated with chronic exposure of T cells to viral antigen, T cells eventually lose the ability to produce IFN-γ. 55,56,65,66 HIV-specific CD8 T cells with elevated levels of PD-1 also show greater susceptible to apoptosis that was attributed to lower levels of the pro-survival Bcl-2 and higher levels of CD95/Fas surface receptor compared to the PD-1 low T cells. 67 The focus of T-cell exhaustion is often on the loss of CD8 T cells function that is primarily responsible for the killing infected cells.
However, CD4 helper T cells also exhibit functional defects during HIV infection. Exhausted CD4 T cells exhibit a reduced HIV-specific proliferative capacity and a loss in polyfunctional cytokine response that centers on reduced IL-2 production. 68,69 Exhausted virus-specific CD4 T cells also express PD-1 with elevated levels correlating with disease progression, viral loads and reduced CD4 T-cell count. 70 As such, PD-1 is a common regulator of exhaustion on HIV-specific CD4 and CD8 T cells. In contrast, the CTLA-4 ICI is more selectively upregulated on exhausted CD4 T cells that correlates with disease progression and T-cell dysfunction. 71 Conversely, the ICIs 2B4 and CD160 that are characteristically upregulated on exhausted CD8 T cells are virtually absent from exhausted CD4 T cells. 72 As with CD8 T cells, expression of multiple ICIs including PD-1, CTLA-4, and TIM3 is associated with a more pronounced state of functional CD4 T-cell exhaustion. 73 A characteristic feature of CD4 and CD8 exhausted T cells is both experience at least partial restoration of antigen-specific proliferative and functional activity following antibody-mediated ICI blockade therapy.
Aside from increased levels of ICIs during HIV infection, an activated T-cell phenotype with upregulated levels of CD38 on CD8 and CD4 T cells is a well-established predictive marker for disease progression. 74 Additional activation markers upregulated during HIV infection include CD38/HLA-DR coexpression on CD8 T cells and signs of ongoing replication as determined by Ki-67+ T cells. 75 T cells also exhibit impaired T-cell maturation characterized by reduced expression of CD28 and high levels of CD27 costimulatory molecules, suggesting a decreased effector phenotype. 76 Decreased levels of CD28 on HIV-specific CD8 T cells are associated with shorter telomere lengths and reduced proliferation. 77 HIV-specific CD8 T cells with elevated levels of CD27 also have reduced Granzyme A and perforin cytotoxic activity compared to CD27 low effector T cells. 78 Moreover, HIV-specific memory CD8 T cells were also found to have a preterminally differentiated phenotype (CD45RA-CCR7-), when compared to CMV-specific cells that instead expressed a terminally differentiated (CD45RA + CCR7-) phenotype. 79 These studies showed that just prior to increases in patient viremia, PD-1 levels increased on HIV-specific CD8 T cells and these cells exhibited reduced in vitro capacity to kill HIV-infected cells. 85 As such, a delicate balance may exist between T-cell-mediated control of viral infection and the progressive development of T-cell exhaustion.

| T-CELL E XHAUS TI ON IN OTHER VIR AL INFEC TIONS
Pathogens have evolve through natural selection to evade immune- HBV and HCV infection where depletion of CD8 cells lead to a prolonged viremia that only declined when CD8 T cells returned. 86,87 The two primary mechanism for chronic hepatitis infection in patients is viral mutation that leads to escape from antiviral CD8 T cells and through exhaustion where CD8 T cells lose their effector function. A strong and early CD4 T-cell helper response in HCV infection is also associated with viral clearance in patients. In contrast, development of HCV-specific CD4 T cells with limited proliferative potential at an early stage following infection resulted in the evolution of a chronic infection. 88

| MA JOR S I G NALING PATHWAYS INVOLVED IN T-CELL E XHAUS TI ON
T-cell activation takes place in a highly organized process where the antigenic peptide-MHC complex on APCs binds to its cognate T-cell TCR complex forming the core of the cell-cell interaction. This primary contact sets into motion the reorganization of cell surface and cytosolic molecules leading to the formation of the immunological synapse. It is during this contact with the APC that ICIs exert their inhibitory effect to suppress signaling events needed for a fully activated T-cell response. Formation of the full catalytically activation AKT, phosphorylated at S473 by mTORC2, was also inhibited in the exhausted T cells (Figure 4). The structure of human PD-1 (hPD-1) in complex with the anti-PD-1 NB01 Fab was solved by molecular replacement using crystals that diffracted to 2.2 Å resolution (6HIG). Molecular modeling by Cα superpositioning of hPD-1 coordinates with the pembrolizumab (PDB 5GGS) and the nivolumab (PDB 5GGR) confirms that NB01 Fab binding to PD-1 does not interfere with the binding of either pembrolizumab or nivolumab anti-PD-1 Abs. The hPDL-1-binding surface (PDB 4ZQK) on hPD-1 is colored in purple and is distinct from the binding epitope of the NB01 Fab that is non-blocking of the PD-1/PDL-1 interaction a binding site for a yet to be identified alternate PD-1 ligand or a region important for transmitting the negative regulatory effect of PD-1.

| RECOVERY OF T-CELL FUN C TI ON: EFFEC T OF ART
Upon initiating ART, the majority of patient have a dramatic reduction of viral load and HIV-1 productively infected cells in both the periphery and lymph nodes. 111 However, despite a spectacular efficacy in reducing morbidity and mortality associate with HIV infection, these drugs only inhibit viral replication and cannot cure those infected. ART maintains the level of plasma HIV-1 RNA below the limit of detection for most treated patients. However, long-lived latently infected memory CD4 T cells persist and the potential for residual virus replication prevent the eradication of infected cells. 112

| INFLUEN CE ON I CIS IN PER S IS TEN CE OF HIV
Multiple observational studies have demonstrated a clear association between expression of ICIs and the HIV reservoir. Amongst these, a pivotal study indicated that central memory CD4 T cells expressing PD-1 were enriched in HIV-infected cells, thus providing the first evidence for ICI expression on HIV-infected cells. 123 However, the first demonstration that PD-1 expressing CD4 T cells might also be the source of replication competent virus was provided with the isolation and analysis of PD-1 + and PD-1 hi /Tfh cells from subjects with non-progressive infection and low viremia. 124 Since then, many studies have shown a significant correlation between the frequency of PD-1 + CD4 T cells with HIV persistence during ART in blood 125,126 and in tissues. 127 However, the direct evidence of a clear relationship between HIV reservoir and PD-1 expression came from isolated memory CD4 T cells from blood and lymph node of HIV-infected aviremic ART-treated individuals. These studies demonstrated that inducible replication competent HIV was found to be highly enriched in lymph node PD-1 + CD4 T cells, containing the Tfh cell population. 128 Recently, a further enrichment in HIV infection F I G U R E 6 Anti-PD-1 antibody-mediated restoration of exhausted T cells. (A) In the standard activation model of T cells expressing low levels of PD-1, the immunological synapse forms with the TCR/CD3 complex at the core, surrounded by the CD28 costimulatory receptor within the cSMAC. This distribution forms a close complex of signaling molecules that enhances the T-cell activation cascade. (B) In PD-1 high T cells, PDL-1 binding to PD-1 recruits the PD-1/SHP2 complex into the cSMAC that effectively suppresses signaling through the SHP2 phosphatase-mediated dephosphorylation of TCR/CD3 and CD28 proximal signaling molecules. (C) Use of blocking anti-PD-1 antibodies such as pembrolizumab partially restores T-cell signaling through limiting PD-1/PDL-1-mediated recruitment of the PD-1/SHP2 complex into the cSMAC. The exclusion of SHP2 from the cSMAC reduces dephosphorylation of signaling molecules including ZAP-70, Lck, PI3K, and AKT. However, our studies show that in activated T cells, pembrolizumab binding to PD-1 pulls down a complex that includes SHP2, CD28, and PI3K following T-cell activation. As such, PD-1 may partially suppress T-cell activation by recruiting CD28 away from the cSMAC and suppressing CD28 costimulation through the SHP2 phosphatase. (D) Our newly discovered anti-PD-1 antibody NB01 is non-blocking of the PD-1/PDL-1 interaction and has equivalent antagonistic activity compared to pembrolizumab in restoring T-cell signaling and antigenspecific functional and proliferative activity to exhausted HIV-specific T cells. Based on immunoprecipitation studies, our proposal is that non-blocking anti-PD-1 antibodies act through inhibiting close contact of the PD-1/SHP2 complex with CD28 and associated signaling molecules following T-cell activation. As such, CD28 is free to migrate into the cSMAC and enhance T-cell activation was shown in cells that express multiple ICIs simultaneously including PD-1, TIGIT, and LAG3, suggesting that these inhibitory receptors not only suppress T-cell activation but consequently suppress HIV transcription, 129 and therefore favor HIV latency. 130

| E XPLOITING PD -1 TARG E TING TO PURG E THE HIV RE S ERVOIR
Immunotherapy through PD-1 blockade represents a major breakthrough that has provided a significant clinical benefit to patients for the treatment of different cancers. [135][136][137] In vitro studies using  138 Anti-PD-1 therapy combined with ART vs ART alone in SIV-infected monkeys also had a more rapid suppression of viral loads and delayed rebound after a standardized treatment interruption. 139 Despite the success of these studies and others at boosting the immune-mediated antiviral activity, SIV-infected monkeys were not able to maintain immunological control of the SIV virus. As such, relieving T-cell-mediated exhaustion through anti-PD-1 blockade is unlikely to be successful as a monotherapy. Although results are preliminary for several clinical studies employing PD-1 blockade, the patients tested thus far have only shown a modest response at best. [132][133][134] This indicates that immunotherapy targeting several ICIs in combination with other strategies to reactivate the virus from latently infected cells may be needed to purge the HIV reservoir.
The HIV virus has developed a considerable stealth in evading detection from a patient's immunological response. Antibody-mediated immunotherapy targeting ICIs can address T-cell functional exhaustion.
However, a limitation is the lack of access of HIV-specific cytotoxic CD8 T cells to privileged anatomic compartments including lymphoid organs where persistent viremia and/or residual virus replication may occur in memory CD4 T cells. 140,141 Approaches for the targeted killing of infected cells would provide an orthogonal method of eliminating the highly heterogeneous latent population of infected cells. Passive immunization using broadly neutralizing antibodies (bNabs) against the HIV-1 Envelope protein may contribute to the killing of infected cells through F I G U R E 7 Targeted killing of PD-1-positive HIV-infected cells. PD-1 expressing CD4 T cells represent the primary compartment of HIVinfected cells that produce replication competent virus. Targeted killing of these PD-1 positive cells represents a novel therapeutic strategy that can deplete the HIV reservoir of infected cells. Antibody drug conjugates (ADC) use a toxin-conjugated antibody that binds to the cell surface PD-1 receptor. Internalization of PD-1 results in ADC degradation within the lysosome, resulting in toxin release that specifically kills the PD-1 expressing cell. An alternate strategy for the targeted killing of HIV infected PD-1 positive cells is through antibody-dependent cellular cytotoxicity (ADCC) with an IgG1 anti-PD-1 antibody. The Fc portion of the antibody binds to the FcγRIII receptors expressed on effector cells including natural killer (NK) cells. NK cells release cytotoxic granules that kill the PD-1-positive HIV-infected CD4 T cell antibody-mediated effector function. However, a recent clinical study was unable to show a benefit in reducing HIV-1 persistence in ART suppressed patients with a combined bNab therapy. 142 A strategy currently under evaluation by our group exploits the fact that PD-1 + CD4 T cells from blood and lymph nodes represent the major cell reservoir for replication competent and infectious HIV in chronic and in longterm antiretroviral-treated subjects. An anti-PD-1 antibody drug conjugate (ADC) was developed with a PNU toxin and in vitro studies show specific induction of apoptosis and cell death in PD-1 positive cells.
Anti-PD-1 ADC treatment of CD4 T cells from chronically infected HIV-1 donors significantly reduced viral production relative to a control anti-PD-1 antibody. This therapy was also effective in aviremic ART donors, purging the majority of cells from that were capable of producing infectious virus (Figure 7). 131 Although these data are very encouraging, a primary consideration with all of therapeutic approaches to eliminate the HIV reservoir will be safety. Considerable strides have led to the development of highly potent ART that effectively suppressed viral loads in most patients for decades with limited adverse events and liabilities.
As such, the bar will be set high to demonstrate a clear medical need for the use of curative strategies that present any dangers to the long-term health of patients.

ACK N OWLED G M ENTS
We thank A. Farina and C. Pellaton for their contributions to this manuscript. The Swiss Vaccine Research Institute and the Swiss National Fund (310030_179246) provided funding during the preparation of this manuscript.

CO N FLI C T O F I NTE R E S T
All authors declare no conflict of interest with this review.