The role of cellular proteostasis in antitumor immunity.

Immune checkpoint blockade therapy is perhaps the most important development in cancer treatment in recent memory. It is based on decades of investigation into the biology of immune cells and the role of the immune system in controlling cancer growth. While the molecular circuitry that governs the immune system in general-and antitumor immunity in particular-is intensely studied, far less attention has been paid to the role of cellular stress in this process. Proteostasis, intimately linked to cell stress responses, refers to the dynamic regulation of the cellular proteome and is maintained through a complex network of systems that govern the synthesis, folding, and degradation of proteins in the cell. Disruption of these systems can result in the loss of protein function, altered protein function, the formation of toxic aggregates, or pathologies associated with cell stress. However, the importance of proteostasis extends beyond its role in maintaining proper protein function; proteostasis governs how tolerant cells may be to mutations in protein-coding genes and the overall half-life of proteins. Such gene expression changes may be associated with human diseases including neurodegenerative diseases, metabolic disease, and cancer and manifest at the protein level against the backdrop of the proteostasis network in any given cellular environment. In this review, we focus on the role of proteostasis in regulating immune responses against cancer as well the role of proteostasis in determining immunogenicity of cancer cells.

Proteostasis

Cellular proteostasis is maintained through the integrated action of protein synthesis, folding, and degradation machinery (Fig. 1) (1, 2, 3). Ribosomes and molecular chaperones are directly responsible for the synthesis and folding of proteins (4, 5). Changes in transcription, metabolism, and the cellular environment provide the backdrop against which protein synthesis and stability must be achieved. When proteins have reached the end of their usefulness or have been damaged by environmental factors, they will be degraded by two main pathways: the ubiquitin-proteasome pathway or the autophagy lysosomal pathway (6, 7). Molecular chaperones play a key role in the determination of which proteins will be degraded as well as aiding the process of degradation itself (4, 8, 9).

Figure 1

Overview of proteostasis pathways. 1. Proteins are synthesized on ribosomes, and molecular chaperones aid in synthesis and proper folding. 2. When proteins are misfolded and reach the end of their life or there is a disruption in proteostasis, two major pathways are responsible for protein degradation: A, the ubiquitin-proteasome pathway and B, the autophagy lysosomal pathway. A variety of stresses can promote the formation of misfolded proteins and toxic aggregates. The heat shock response (HSR) is regulated by the transcription factor, heat shock factor 1 (Hsf1). Hsf1 is normally held in an inactive, monomeric state by chaperones such as Hsp70 and Hsp90. When misfolded protein species become sufficiently abundant, Hsf1 can no longer efficiently compete for binding to chaperones and then undergoes trimerization and translocation to the nucleus. The unfolded protein response (UPR) is activated when misfolded proteins accumulate in the ER. The UPR sensors PERK, IRE1, and ATF6 are normally held in an inactive state by the ER chaperone, Grp78. Misfolded proteins titrate Grp78 away from these sensor proteins which are subsequently activated and trigger the UPR. ER, endoplasmic reticulum.

Proteostasis can be disrupted by a variety of extrinsic factors such as nutrient deprivation, thermal stress, hypoxia, oxidative stress, etc., as well as intrinsic factors such as mutations in protein-coding genes that can thermodynamically destabilize their protein products. Cancer cells, and the tumor microenvironments they define, are characterized by intrinsic and extrinsic challenges to proteostasis (10). Cancer cells are heavily dependent on molecular chaperone systems to maintain a functional proteome in the face of an ever-increasing burden of mutations in an environment that is often hypoxic, nutrient deprived, and flush with reactive oxygen species (11, 12, 13, 14, 15, 16). The tumor microenvironment also triggers unfolded protein response (UPR) activation in key immune cells that participate in antitumor immunity (17).

Cellular stress responses have evolved to transiently supply additional machinery to handle proteotoxic stress and avoid cell death. In eukaryotes, the heat shock response (HSR) and UPR are responsible for responding to proteotoxic stress and maintaining proteostasis. The HSR is triggered by cytoplasmic protein misfolding and is induced by the transcription factor, heat shock factor 1 (Hsf1) (18, 19, 20). Hsf1 is normally held in a monomeric state through interactions with chaperones such as Hsp70, TriC, and Hsp90 (21, 22, 23, 24). Elevated levels of misfolded proteins increase demand on chaperones, thereby titrating them away from Hsf1. This allows Hsf1 to undergo trimerization, activation, and translocation to the nucleus where it can act on target genes that aid in the response to protein misfolding (18, 19, 25, 26). The UPR can be induced by at least three different types of stress sensors in the endoplasmic reticulum (ER) (27, 28, 29). The inositol-requiring enzyme (IRE1), the activating transcription factor 6 (ATF6), and the protein kinase R–like endoplasmic reticulum kinase (PERK) all activate the UPR to allow cells a chance to adapt to challenges to proteostasis and modulate the activation of apoptosis (30, 31, 32). Each of these three sensors is activated in response to an elevated burden of misfolded proteins in the ER. IRE1, ATF6, and PERK are normally held in an inactive state by an abundant ER chaperone, Grp78 (also called BiP, Hspa5) (33, 34, 35). When the level of misfolded proteins increases in the ER, these misfolded proteins begin to compete with the stress sensors for binding to Grp78. Sufficient levels of misfolded proteins will eventually lead to the loss of Grp78 binding to PERK, IRE1, and ATF6, leading to their activation and the initiation of the UPR. Release of ATF6 from Grp78 allows for its traffic to the Golgi where it undergoes proteolysis to release a soluble and activate transcription factor domain (cATF6) that can then translocate to the nucleus (36, 37). In the case of IRE1, loss of Grp78 binding to its lumenal domain allows for dimerization, autophosphorylation, and activation of an RNase domain that then completes the splicing of the messenger RNA (mRNA) encoding the transcription factor XBP1 (38, 39). It has also been shown that direct binding of IRE1 with unfolded proteins is important for IRE1 oligomerization and activation (40, 41). Finally, in a manner very similar to IRE1, PERK activation results in autophosphorylation and subsequent phosphorylation of eIF2alpha (33, 42). Phosphorylation inactivates eIF2alpha and results in both a decrease in global translation as well as the translation of the ATF4 open reading frame (ORF) which is normally masked by translation of an overlapping ORF (43, 44).

Key players in antitumor immunity

The immune system plays a critical role in detecting and eliminating cancer cells from the body (45, 46). However, it was not until the 1980s that the definitive link between the immune system and cancer was made despite the idea being proposed in the 19th century (47, 48, 49, 50, 51). This was in part owing to the difficulty associated with impairing specific immune cell lineages in animal models. Two of the most important immune cells for antitumor immunity are a type of antigen-presenting cell (APC) called dendritic cells (DCs) and cytotoxic T cells (CTLs) (Fig. 2A).

Figure 2

Antitumor immunity and immune checkpoint blockade therapy. 1. Tumor-derived material is phagocytosed, processed, and presented by antigen-presenting cells (APCs) such as dendritic cells (DCs). 2. DCs presenting “nonself” peptides will activate cognate T cells. A, activated T cells will recognize MHCI complexes loaded with neoantigenic, nonself peptides on T cells, resulting in tumor cell killing. B, expression of PD-L1 on tumor cells will block T-cell action. C, neutralization of the PD-1–PD-L1 interaction with therapeutic antibodies will restore T cell–mediated killing. ER, endoplasmic reticulum; MHCI, major histocompatibility complex type I; PD-1, programmed death 1; PD-L1, programmed death ligand 1; PLC, peptide-loading complex; TAP, transporter associated with antigen processing.

DCs infiltrate the tumor in response to innate immune signaling such as inflammation (52, 53). These tumor-infiltrating DCs internalize material from dead cancer cells or tumor-derived extracellular vesicles, the protein components of which are then degraded into fragments for loading onto major histocompatibility complex type I (MHCI) and II (54, 55, 56). These complexes then translocate to the surface of the APC where they are “presented” to CTLs or helper T cells, respectively, in nearby lymph nodes (52). T cells that can recognize specific nonself antigens presented on the surface of these DCs will become activated and then launch an immune response against that particular epitope or antigen, in the tumor (57, 58). Normally nonself antigens are derived from pathogens, such as viral proteins, but in the context of cancer, peptides are derived from mutated gene products. Since all cancer cells possess mutations that encode potentially neoantigenic proteins, how can cancer cells escape the immune system at all?

Immune evasion via programmed death ligand 1

All nucleated vertebrate cells express MHCI on their surface that present peptide fragments derived from protein degradation for surveillance by the immune system. Unlike in APCs like DCs, which present antigens to activate immune cells, the presentation of neoantigens in MHCI by cancer cells renders them vulnerable to attack by CTLs in a manner that is often proportional to their mutagenic burden (59, 60). Complete loss of MHCI seems like an obvious mechanism for immune evasion, but this renders cancer cells vulnerable to killing by natural killer (NK) cells owing to the role that MHCI complexes play in inhibiting these cells (61, 62, 63, 64). However, the relentless evolution of new traits in tumors allows them to evade immune detection and alter immune cell function in the tumor microenvironment. The most important example of this is the transmembrane protein, programmed death ligand 1 (PD-L1) (65, 66). PD-L1 binds to another transmembrane protein called programmed death 1 (PD-1) that is expressed on CTLs (Fig. 2B) (67). PD-L1 is normally expressed by some macrophages, act9vated T and B cells, and DCs to maintain a balance between activation, tolerance, and immune-mediated tissue damage (68). Tumor cells express PD-L1 to protect them from CTL-mediated killing and are the focal point of immune checkpoint inhibition (ICI) (69, 70, 71, 72, 73). Antibodies directed against PD-1 (e.g., nivolumab, pembrolizumab) or PD-L1 (e.g., avelumab, atezolizumab) neutralize the PD-1–PD-L1 interaction, restore CTL-mediated killing, and lead to tumor regression in many settings (Fig. 2C) (74, 75, 76, 77, 78, 79).

The expression of immunomodulatory transmembrane proteins like PD-L1 can be influenced by different facets of the proteostasis network. Grp78 is an ER-resident member of the Hsp70 chaperone family that is important for the stability and trafficking of PD-L1 (80). Upregulation of Grp78 by the UPR is one mechanism through which tumors can promote the surface localization and accumulation of PD-L1 (81). Other chaperones have also been implicated in the regulation of PD-L1 stability and surface levels. Mammalian cells express four different Hsp90s (Hsp90ɑ and Hsp90β in the cytoplasm, Trap1 in the mitochondria, Grp94 in the ER) that can be inhibited by various highly specific pharmacological agents (82, 83). Inhibition of Hsp90 with ganetespib has been shown to decrease surface levels of PD-L1 in a mouse colon cancer cell line (84). However, another study examining the same cell line, albeit with a different Hsp90 inhibitor and at a lower concentration, did not show a change in PD-L1 levels but did result in an increase in MHCI complexes (85). These different studies may indicate that there may be different thresholds for changes in immunomodulatory systems with respect to proteostasis.

Immune evasion via changes in peptide generation

The proteasome is an important part of proteostasis, regulating the degradation of unneeded and damaged proteins. The 26S proteasome is responsible for the degradation of these polyubiquitinated proteins (86, 87). Polyubiquitinated proteins arise from the cascade of E1, E2, and E3 enzymes, which activate, conjugate, and transfer ubiquitin moieties to protein substrates (88, 89). The proteasome is a multisubunit complex that is composed of a 20S catalytic core and capped by two 19S regulator complexes on each end (90). The 20S core is composed of four stacked rings of seven subunits each, with seven α-subunits (α1–α7) forming the two outer rings and the two inner rings are composed of 7 different β-subunits (β1–β7) (91, 92). There are three specialized proteolytic subunits found in the heptameric β rings of the 20S core, β1, β2, and β5, which show different cleavage preferences. The proteasome produces peptides that are 8 to 10 residues in length, providing peptides that can be accommodated by the MHCI peptide-binding groove. Inhibition of the proteasome reduces peptide production and results in significantly reduced MHCI antigen presentation (93, 94). Interestingly, several of the components of the MHCI antigen presentation pathway are induced by the cytokine interferon-γ (IFNγ) (95, 96). All three of the β subunits with proteolytic activity have counterparts that are IFNγ inducible and define the so-called immunoproteasome. Specifically, immunoproteasomes contain the immunosubunits iβ1, iβ2, and iβ5 (also known as PSMB9/LMP2, PSMB10/MECL1, and PSMB8/LMP7, respectively) in place of the β1, β2, and β5 subunits contained in the standard proteasome. While both the standard proteasome and immunoproteasome are able to generate peptides that can be loaded onto MHCI complexes, immunoproteasomes have altered cleavage preferences that result in peptides with hydrophobic C termini and a different cleavage rate that support immune response (97, 98, 99, 100, 101). Peptides derived from immunoproteasomes are far more efficiently loaded into the MHCI peptide-binding groove. Mouce models show that deficiency in the immunoproteasome can reduce T-cell response and altered antigenic peptide presentation (102, 103, 104, 105). This difference in proteasome subtypes was again demonstrated in HeLa cells infected with hepatitis B virus, where only after stimulation with IFNγ did cells present hepatitis B virus epitopes (106, 107).

As the presentation of neoantigens by cancer cells is important for immune cell recognition, it makes sense that malignant cells suppress immunoproteasome function in order to avoid immune detection (108, 109). Low immunoproteasome expression in patients with non-small cell lung cancer is a prognostic marker for reduced survival and increased recurrence of metastases (110). Cancer cells need to limit immune recognition linked to peptide presentation while also relying heavily on the proteasome to maintain proteostasis due to protein imbalances caused by global expression changes (i.e., linked to aneuploidy, gene amplification, overexpression, mutation, and hyperactivation of oncogenes). Proteasome inhibitors, such as Velcade, have been in use for cancer treatments for almost 20 years as a way to induce apoptosis and cell death through a variety of different molecular effects (111, 112). While proteasome inhibition has been shown to alter peptide presentation, the two are not necessarily synonymous (93, 113, 114). Treatment of cells with the proteasome inhibitors has also been shown to increase the levels of intracellular peptides, presumably by shifting the task of degradation to different proteolytic systems (115, 116).

A majority of nonproteasomal peptidases are thought to play supportive roles to the proteasome, performing further trimming and degradation. While these pathways can ultimately lead to the destruction of peptides, there is evidence that both in conjunction with the proteasome and independently these peptidases play important roles in the processing of antigenic peptides and immune cell recognition (117, 118). Similar to the proteasome and other members of the proteostasis machinery, the expression of alternative peptidases has been shown to be induced in cancer and is often correlated with disease outcomes (119, 120, 121). The precise mechanisms through which these alternate proteolytic mechanisms become altered in cancer and how they can affect MHCI peptide trimming and loading are not known. However, one can speculate that shifts in the immunopeptidome, and resultant immunogenicity, could be linked to alternative proteolytic pathways.

One of the most fundamental players in regulating the targeting of proteins to degradation pathways are chaperones tasked with folding, and maintaining, proteins in the cells. As mentioned earlier, inhibition of Hsp90 has been shown to have varying effects on the stability and trafficking of PD-L1 in cancer cells (84, 85). In one of these studies, mild Hsp90 inhibition—that did not result in the induction of the HSR or UPR or changes in PD-L1 expression or localization—promoted the formation of MHCI complexes and their accumulation at the cell surface (85). This suggests that Hsp90 inhibition promotes the targeting of Hsp90 client proteins to the ubiquitin-proteasome pathway, has an effect on the MHCI assembly pathway, or both. Interestingly, this study showed in animal models that this mild Hsp90 inhibition resulted in an immune system–dependent reduction in tumor growth (85). This, together with their observation that mild Hsp90 inhibition resulted in changes in the immunopeptidome, suggests that chaperones like Hsp90 can have a profound effect on the immunopeptidome of cancer cells.

Immune evasion via changes in peptide transport, MHCI assembly, and peptide loading

The assembly of MHCI molecules with peptide antigens involves the interaction and assembly of multiple components (Fig. 3). The sequential formation of the different interactions is difficult to dissect; however, several key interactors assist in peptide loading. Tapasin is an MHCI-specific chaperone that binds to peptide receptive MHCI heterodimers and to transporter associated with antigen processing (TAP), connecting MHCI to the source of peptides (122). ERp57 is a thiol oxidoreductase that functions together with the ER chaperones calnexin and calreticulin to stabilize MHCI and to stabilize the interaction of MHCI, calreticulin, tapasin, and TAP together as the peptide-loading complex (PLC) (123, 124). Tapasin plays a key role in bridging the components of the PLC together (125, 126).

Figure 3

MHCI antigen presentation pathway. MHCI heavy chain is stabilized by the chaperone calnexin until it is replaced by its soluble counterpart, calreticulin, which helps to stabilize the binding of β2-microglobulin (β2m). This complex together with ERp57, Tapasin, and the TAP transporter form the peptide-loading complex (PLC). Peptides are transported through TAP and loaded onto MHCI. Kinetically stable MHCI–peptide complexes are released from the PLC and move through the secretory pathway to the cell surface for surveillance by cytotoxic T cells. ER, endoplasmic reticulum; MHCI, major histocompatibility complex type I; PD-1, programmed death 1; PD-L1, programmed death ligand 1; TAP, transporter associated with antigen processing.

Structural modulations and dysregulation of the components of the PLC can alter MHCI surface expression and change disease outcomes (127, 128, 129, 130). As an MHCI-specific chaperone, changes to tapasin expression can be directly linked to reduced MHCI antigen presentation. Loss of tapasin in cancer cells results in escape from recognition by CTLs and low levels of tapasin have been correlated with poor prognosis and patient survival time (131, 132, 133). In a similar way, the downregulation of TAP1 has been shown as a mechanism of tumor immune escape and implicated in tumorigenicity in several cancers (134, 135, 136, 137). Additional members of the PLC have more general roles in cellular function, and therefore, a direct line between MHCI antigen presentation and their regulation is more difficult to define. The ER chaperones calreticulin and calnexin are core components of the ER quality control pathway and are upregulated as part of the UPR (138). Induction of the UPR has been demonstrated to reduce expression of MHCI on the cell surface (139, 140). Reduction of MHCI levels due to the UPR has also been shown to promote NK-mediated killing in thyroid cells (141, 142, 143). In contrast, downregulation or mutation of calreticulin also decreases MHCI expression on the cell surface (144). In addition to a role in MHCI stability, calreticulin also functions in the folding and stabilization of tapasin and can bind to MHCI independent of the PLC (145, 146). In calreticulin-deficient cells, tapasin levels were reduced, indicating that calreticulin may influence MHCI complex formation both directly and indirectly (147). All of this suggests that the balance of chaperones in the ER is important to determine how efficiently MHCI complexes will form and how they will traffic to the cell surface. In addition to lower levels of MHCI complexes on the surface of cells when calreticulin activity is compromised, many of these complexes are suboptimally loaded with peptides. Interestingly, calreticulin plays an important role in the retrieval of such suboptimally loaded MHCI from post-ER compartments (148, 149). Loss or reduction of this retrieval function could explain the reduced MHCI antigen presentation observed in calreticulin-deficient cells.

Calreticulin itself can also be expressed on the cell surface as a marker for immunogenic cells (150). Levels of calnexin are known to alter antigen presentation and the cancer immune response (151, 152). Interestingly, the downregulation of calnexin was recently shown to be directly associated with lower expression of MHCI and enhanced cancer progression (153). Hsp90 has also been implicated in the loading of peptides into MHCI. Inhibition of Hsp90 has been shown to result in the appearance of unloaded MHCI on the surface of some cancer cells lines (154). To what degree alterations in proteostasis resulting in the appearance of unloaded MHC complexes at the cell surface of cancer cells influences recognition by the immune system remains to be examined.

Human leukocyte antigen expression and cancer

The type of MHCI complex on a cancer cell can also be as important for CTL-mediated killing as the peptide that is loaded within it. While the beta chain of a typical MHCI complex is always β2 microglobulin, there are several alpha chain genes, encoding different human leukocyte antigens (HLA), that can be expressed (155). In addition to the three classical alpha chains (HLA-A, HLA-B, and HLA-C in humans), there are three nonclassical HLA genes as well (HLA-E, HLA-F, and HLA-G). All of these HLA heavy chains differ in their peptide binding preferences and expression. Moreover, there are thousands of known alleles of HLA-A, HLA-B, and HLA-C and dozens of alleles of HLA-E, HLA-F, and HLA-G, giving each person a unique HLA profile and corresponding immunopeptidome in each cell type. HLA-A and HLA-B are the most well studied and have different peptide binding preferences (156). For example, HLA-A binds selectively to hydrophobic peptides, while HLA-B can bind to peptides with a wider range of properties. The nonclassical HLA-E binds to a limited number of peptides, usually derived from the signal sequences of other HLA molecules (A, B, C, and G). On the cell surface, these HLA-E MHCI complexes (loaded with MHCI signal peptides) can inhibit CTLs and NK cells (157, 158). Interestingly, the mitochondrial import sequence of Hsp60 can also be loaded onto HLA-E complexes and traffic to the cell surface (159). However, HLA-E complexes loaded with Hsp60-derived peptides do not inhibit NK cells and may provide a means of promoting immune clearance of stressed cells. It is intriguing to consider the possibility that the significance of HLA-E expression in disease severity can only be determined against the backdrop of the UPR status of the tumor. It may be that higher levels of stress, and concomitant expression of Hsp60, may cause a switch from an immune protective role to one that attracts NK cells. Another nonclassical MHCI, HLA-G, is often overexpressed in cancer and has been shown to facilitate immune evasion in melanoma (160). It has been proposed that HLA-G may constitute a novel immune checkpoint (161, 162). Similarly, high expression of the nonclassical MHCI, HLA-F, is associated with poor prognosis (163, 164).

Cancer severity, and response to ICI, can be influenced in a variety of ways by HLA genetics and expression (165). Clear links between proteostasis and HLA expression profile have yet to be thoroughly investigated, let alone established, but there is good reason to pursue such lines of research. An excessive burden of MHC complex loading and folding can induce the UPR, and induction of the UPR can influence MHCI surface expression (141, 166). Taken together, it seems reasonable that proteostasis can play a role in determining which MHCI complexes reach the cell surface and what effect they have on tumor immunogenicity once they arrive.

Future prospects

The immune system may be our best ally in the fight against cancer. Some of the most dramatic treatment responses to immunotherapy have been seen in melanoma where the combination PD-L1 and CTLA-4 (another inhibitory transmembrane protein on T cells that normally prevents hyperproliferation and autoimmunity) blockade showed an overall median survival of longer than 60 months and a five-year overall survival of 60% (78, 167, 168). In non-small cell lung cancer, treatment with PD-1 blockade resulted in a significantly increased 4-year survival compared to chemotherapy (14% vs 5%) (169). Unfortunately, despite the tremendous advancements in immunotherapy, it is estimated that checkpoint inhibitors may only benefit approximately 13% of all patients with cancer (170). The reasons for why many tumors are “cold” to the immune system and thus do not respond to ICI is not well understood but is currently a major research focus (171, 172, 173). A cold tumor is often characterized by having little to no infiltration by immune cells or in which immune cells accumulate at the periphery of the tumor (174). “Hot” tumors are characterized by intense interferon gamma signaling, inflammation, and T-cell infiltration which in turn usually correlates with a positive response to ICI. While this is a general definition of cold and hot tumors, it is not absolute. A recent multiomic analysis of melanoma did not reveal a clear set of predictive elements for response to immunotherapy (175). Some tumors that appeared cold responded well to immunotherapy, and some tumors with all the hallmarks of a hot tumor did not. More focused analysis on cellular stress responses and proteostasis in the context of immunotherapy may lead to a more accurate definition for hot and cold tumor and lead to better strategies to treat patients.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.