Immune checkpoint inhibitors:143 New generation of ICI and stimulatory approaches based on the emerging scientific insights summarized above are now in various phases of clinical development and will likely play a pivotal role in combatting resistance mechanisms to current ICI (table 2).144–146

Primary versus secondary resistance

As the majority of tumors are resistant to ICI147 there has been considerable effort focused on identifying mechanisms of ICI resistance and developing subsequent therapeutic solutions. In 2019, SITC developed definitions for both primary and secondary (or adaptive) resistance to PD-1 pathway blockade to promote semantic consistency in both clinical and research endeavors.148 The taskforce defined primary resistance as disease progression after at least 6 weeks but no more than 6 months of anti-PD-(L)1 therapy.148 Secondary resistance was defined as disease progression following an objective or prolonged (>6 months) response to PD-(L)1 blockade.148 Primary resistance mechanisms include certain oncogene-driving mutations, both lack of quantity and/or quality of neoantigen presentation, and the metabolically hostile and toxic TME that impedes T-cell function.149 150 Secondary resistance can be attributed to the upregulation of coinhibitory signaling, selective loss of antigenic tumor populations via immunoediting, and T-cell exhaustion.149 151 However, given the large degree of overlap between primary and secondary resistance mechanisms, further description could be more clearly grouped into tumor intrinsic versus extrinsic factors.

Review of key mechanisms of resistance to checkpoint blockade

Tumor immunogenicity plays a key role in the response to ICI as it serves as the source of pre-existing adaptive immune response. Several biomarkers have been developed to quantify the neoantigen load of each tumor. TMB is a widely used surrogate for tumor neoantigenicity, though not all cancers exhibit a clear correlation between TMB and ICI efficacy.152 153 MMR-deficient and MSI-high tumors exhibit increased insertion-deletion (indel) mutations that more strongly correlate with ICI benefit, prompting a tumor agnostic Food and Drug Administration approval of pembrolizumab for this indication.154 However, these biomarkers only scrape the surface in characterizing an individual tumor’s neoantigen load and immunogenic potential, thus highlighting the chasm in our understanding of ICI resistance.155 Fortunately, advances in predictive modeling may provide more accurate biomarkers of “true neoantigen load” by using novel major histocompatibility complex (MHC)-binding site mutations, TCR recognition features, and intratumoral heterogeneity.156–158 Regardless, selective pressure will inherently lead to an immunoediting process by which the less immunogenic subpopulations will outcompete and lead to cancer progression.159 Estimation of the immune-editing combining whole-exome data (ie, lower number of observed vs neoantigens), gene-expression data, or IHC assessment, is an emerging field of research.160–162

Specific oncogene signaling and other mutations are also intrinsic elements of tumor immune escape. For example, mutations in KRAS, EGFR, and MAPK not only promote unchecked cellular growth, but can also lead to poor anti-PD(L)1 response due to repressed IRF2 expression, myeloid-derived suppressor cell (MDSC) recruitment, decreased PD-L1 expression, and decreased T-cell infiltration, respectively.163–165 Additionally, loss of PTEN signaling recruits immunosuppressive cells to the TME while dysregulation of the Wnt/β-catenin cascade has been shown to impair naïve T-cell priming in melanoma.166 167 Finally, mutations in the JAK1/2 pathway can disrupt the antitumor effects of IFN-γ signaling post-MHC-TCR interaction while mutations in the β2 microglobulin component of MHC-I molecules can prevent neoantigen loading and presentation.168 169 These data again highlight that the quality of TMB may be more significant than pure quantity as measured by current TMB assays.

Based on these genomic and transcriptomic alterations, several novel agents have now entered clinical investigation. For example, numerous Wnt/β-catenin pathway inhibitors have shown efficacy in preclinical models, one of which (CGX1321) has been combined with pembrolizumab or targeted therapy in a trial of patients with pretreated advanced gastrointestinal tumors. Sotorasib, a KRASG12C inhibitor, has recently shown encouraging results in the clinic: CodeBreak 100/101 showed durable efficacy and an acceptable safety profile of lead-in sotorasib followed by anti-PD(L)1 therapy in patients with pretreated NSCLC, and other cohorts are under investigation (NCT04185883).170 Further, the ubiquitous MAPK oncogenic pathway has now been targeted with a p38-specific inhibitor, ARRY614. A phase Ib trial is now assessing ARRY614 in combination with ICI in treatment refractory advanced solid tumors (NCT04074967).

In recent years, the TME has been shown to harbor several immunosuppressive features tied to ICI resistance, including the upregulation of novel checkpoint receptors, recruitment of regulatory and suppressive cell populations, concentration of toxic metabolites, and dysfunction and terminal exhaustion of effector T cells.149 171 As noted, a cascade of clinical trials is underway to assess the efficacy of novel co-inhibitory receptor antagonists alone or alongside anti-PD-1 therapy—including LAG-3, which was recently approved in melanoma in combination with nivolumab, TIM-3, TIGIT, VISTA, BTLA, and ICOS inhibitors, among others (figure 2a).172 Given preclinical data showing upregulation of both TIM-3 and LAG-3 following anti-PD(L)1 therapy, with dual blockade decreasing tumor-infiltrating lymphocytes exhaustion and tumor progression, combinatorial checkpoint inhibition may be a key route toward improving immune responsiveness.173 174 Recently, the addition of tiragolumab (anti-TIGIT) to atezolizumab in patients with PD-L1+ metastatic NSCLC did not show significantly improved PFS versus anti-PD(L)1 alone in the front-line setting phase III SKYSCRAPER-01 trial, indicating the need for more nuanced predictive biomarkers for combinatorial checkpoint therapy (NCT04294810). Finally, in addition to co-inhibitory agents, numerous co-stimulatory receptor agonists have also entered clinical trials to overcome resistance mechanisms in the TME. These targets include 4-1BB, OX40, ICOS, GITR, CD40(L), and CD27, many in combination with PD-1 blockade (figure 2a).172 The recently reported first-line phase II PRINCE trial combining sotigalimab (CD40 agonist) and/or nivolumab with chemotherapy in patients with advanced pancreatic adenocarcinoma identified blood and tumor-based biomarkers of survival after sotigalimab plus chemotherapy and nivolumab plus chemotherapy, whereas a patient subset benefitting from sotigalimab plus nivolumab plus chemotherapy was not identified. This study suggests that biomarker-based patient selection for combination treatments to enhance efficacy may be needed in future trials of chemoimmunotherapy combinations.175

Figure 2

(a) Co-inhibitory (red) and co-stimulatory (green) receptors of T-cell activation in the tumor microenvironment; agents targeting these molecules are currently under various phases of clinical investigation. (b) Innate immune targets in the tumor microenvironment currently under various phases of preclinical and clinical investigation. NK, natural killer; TAM, tumor-associated macrophage.

While CD8+T cells are often the focus of ICI response, multiple immunosuppressive cells have been shown to hamper antitumor immunity and promote ICI resistance, including regulatory T cell (Treg) cells, MDSCs, and tumor-associated macrophages (TAMs).176–178 For example, selective targeting of MDSCs via cytokine signaling was shown to improve anti-PD-1 responsiveness in a preclinical CRC study, and inhibition of the TAM-dependent CSF1R protein was shown to improve ICI treatment in a pancreatic cancer model179 180 (figure 2b). Antagonists of the CCR5-CCL5 axis, CSF1R, and CXCR2 have now entered clinical trials with or without checkpoint blockade in treatment refractory advanced solid tumors.181 Furthermore, several recent preclinical and translational publications have demonstrated the interaction between TAMs, adaptive immunity, and how TAMs suppress adaptive immunity in depth with novel mechanisms.182–185 A recent article demonstrated a spatiotemporal co-dependency between TAMs and CD8+exhausted T cells.182 On the other hand, several publications demonstrate that distinct populations of TAMs have antitumor function. A subset of TAMs C1qa+macrophages have been shown to promote T effector cells by upregulating m6A methyltransferase Mettl14 and epigenetic modulation.183 Mettl14-deficient C1q+TAMs show a decreased m6A abundance and a higher level of transcripts of Ebi3, a cytokine subunit that suppresses T-cell function.183 In patients with breast cancer, a subset of tissue residential macrophage FOLR2+that exists in both healthy and tumor tissue correlates positively with CD8+T cells and antitumor immunity.184 In contrast, TERM2+macrophages are predominantly in tumor but not healthy tissue.184 185 In addition, targeting TREM2+macrophages with anti-TREM2 antibody sensitizes to PD-1 treatment.185 Therefore, it is critical to investigate macrophage heterogeneity in different types of tumors and develop subset-targeted therapeutic strategies for macrophage-based cancer therapies. Strategies to target myeloid cells, including TAMs to enhance immunotherapy, are currently under active clinical investigation. Blockade of CD47-SIRPa axis to enhance phagocytosis of macrophages by antagonizing the “don’t eat me” signal in a tumor is widely studied as a novel means to promote antitumor immunity.186 187 Rational combination of CD47-blockade with treatments that increase prophagocytic marker have yielded some promising results. For example, magrolimab plus rituximab has shown 50% ORR and 36% CR in relapsed/refractory lymphoma.188 In addition, magrolimab combined with checkpoint therapy showed promise in advanced solid tumors.189 cGAS-STING signaling in macrophages, responsible for sensing cytosolic dsDNA, plays a critical role in antitumor immunity by inducing the production of type I IFNs.190 Therefore, STING agonists are actively being studied as a novel immunotherapeutic strategy, with multiple ongoing trials, including in combination with checkpoint blockade.191 192 Further, toll-like receptors (TLRs) are key regulators of macrophage function and differentiation. TLR activation can drive macrophage towards a more “M1-like” pro-inflammatory phenotype.193 In a recent phase II study, intratumoral injection of CMP-001—a virus-like particle containing a TLR9 agonist—in combination with PD-1 blockade, provided local and distant responses in patients with advanced anti-PD-1 resistant melanoma (figure 2b).194 A phase II neoadjuvant trial of CMP-001 with ant-PD-1 in patients with melanoma is currently underway (NCT04401995). Trials of TLR 7/8 receptors, which have the potential to activate a broader range of APCs versus targeting TLR9 within the TME, are also underway.195 However, certain strategies targeting macrophages, such as pattern recognition receptors, have not demonstrated success in large clinical trials. Moving forward, it will be critical to identify the underlying mechanisms of resistance to agents targeting macrophage/myeloid cell biology using clinically derived samples, which will further inform novel combination strategies.

Reversing metabolic insufficiency in the TME to improve outcomes with ICI is also an area of great interest. Tumor-infiltrating T cells face increased acidity, hypoxia, oxidative stress, dysregulated amino acid and lipid metabolism, all of which inhibit antitumor immunity.171 Early phase clinical trials have now targeted the lactate shuttle via MCT1 inhibition in advanced solid cancers, and metformin has been combined with PD-1 blockade to reverse hypoxia and T-cell dysfunction in the TME.171 As dysregulated tumor cell metabolism is associated with hypoxia in the TME,196 targeting tumorous vasculature to remodel the TME is another area of active investigation, including combinations of anti-PD-1 and anti-angiogenic therapy across multiple indications,197 such as anti-PD-1 refractory melanoma.198 199 Primarily through HIF-1-dependent processes, hypoxia also promotes adenosine-mediated T-cell dysfunction and dysregulated neovascularization. Thus, numerous clinical trials targeting the adenosine pathway (eg, CD39, CD73, A2AR) are now underway in combination with anti-PD-1 therapy.200 201 As previously discussed, obesity and the microbiome have been linked to differential ICI response, prompting investigation into modulators of lipid metabolism and fecal microbiota transplant to augment checkpoint blockade.124 202 While most resistance mechanisms ultimately foster T-cell exhaustion, the lack of anergy alone is insufficient for ICI response. Multiple studies have shown significantly increased levels of memory CD4+ T cells in anti-PD-1 responding patients, highlighting the role of T-cell differentiation in acquired resistance.203 204 With advances in single-cell and TCR sequencing, autologous TIL therapy may be further refined for the expansion of potent, highly selective antitumor lymphocytes—a critical development as upcoming trials specifically target the ICI-resistant population.205 206 Box 4 summarizes take home messages and unanswered questions regarding the impact of therapeutic interventions via checkpoint inhibitors to CIR.

Box 4 Take home messages for therapeutic interventions—checkpoint inhibitors

Single and dual agent immune checkpoint inhibition (ICI) has been approved across a wide range of malignancies and disease indications, including in the neoadjuvant, adjuvant, and metastatic setting.

Many patients do not derive clinical benefit from therapy, and intense efforts are underway to elucidate mechanisms of resistance.

Primary resistance is defined as disease progression after at least 6 weeks but no more than 6 months of anti-PD-(L)1 therapy, whereas secondary resistance is defined as disease progression following an objective or prolonged (>6 months) response to PD-(L)1 blockade.

There is a large degree of overlap between primary and secondary resistance mechanisms. These include reduced tumor neoantigenicity, oncogene signaling and mutations driving the tumor microenvironment (TME) towards an immunosuppressive phenotype, upregulation of novel checkpoint receptors, recruitment of regulatory and suppressive cell populations, and concentration of toxic metabolites resulting in a metabolically harsh TME, all of which result in dysfunction and terminal exhaustion of effector T cells.

The next generation of ICI and stimulatory approaches are now in various phases of clinical development and will likely play a pivotal role in combatting resistance mechanisms to current ICI.

The most successful strategies will likely target non-redundant mechanisms of within the cancer immunity cycle.

Unanswered questions for therapeutic intervention contributions to cancer immune responsiveness

Can predictive biomarkers be characterized towards determining patients who are unlikely to benefit from immunotherapy, and for whom other therapeutic strategies be pursued up-front?

How can we best identify patients that may require dual checkpoint blockade versus anti-PD-(L)1 monotherapy?

Can the field identify predictive biomarkers towards characterizing patients that are likely to have significant toxicity so that treatment may be de-escalated and/or IO sparing strategies may be used?

Given the inherent issues of intra and intertumoral heterogeneity, how can we confidently determine which combination strategy to move forward with in each patient?

Several agents have low to minimal single-agent clinical efficacy but have shown clinically meaningful results in combination. How can the field confidently move towards development in combination of drugs with limited efficacy as monotherapy?

Given the failure of later phase trials following promising phase II results, how can we best determine the most rational combinations to proceed with in later phase trials?