Based on a subgroup of the KEYNOTE-001 trial, in which a small number of EGFRm, TKI-naïve patients experienced improved objective response rate (ORR) compared to TKI-pretreated patients, the follow up Phase II NCT02879994 trial evaluated pembrolizumab first-line therapy in TKI-naïve, EGFRm advanced NSCLC patients with PD-L1 positivity (TPS ≥ 1%) (38, 58). This trial was halted early due to futility, as only 1/11 initial patients experienced an OR and this patient was found on subsequent analysis to be EGFRwt. Importantly, 6/7 patients that switched to second line TKI therapy after PD on pembrolizumab experienced treatment-related adverse events (TRAEs) from the TKI (erlotinib), including one patient who experienced grade 5 pneumonitis. These results suggested potential toxicity of EGFR TKI therapy after pembrolizumab treatment in EGFRm NSCLC patients.

A small subgroup of patients in the CheckMate 012 study examining first-line nivolumab monotherapy were EGFRm (6/56, 11%). The ORR for EGFRm versus EGFRwt patients was 14% versus 30%, indicating a comparatively decreased efficacy of ICI therapy in the first-line for EGFRm patients (37).

The CheckMate 012 trial compared niviolumab + platinum-doublet chemotherapy in EGFRwt versus EGFRm advanced NSCLC patients as first-line treatment (39). EGFRm patients experienced worse PFS (4.8 vs 7.5 months) and median OS (20.5 versus 24.5 months) compared to EGFRwt.

The results of the phase 1 NCT02088112 study, a two-part, dose-escalation study with durvalumab + gefitinib as first-line therapy in advanced EGFRm NSCLC patients, were recently published (40). All patients in the study received gefitinib daily. In the dose escalation portion of the study (Part A), 3 patients were randomized to additionally receive durvalumab 3 mg/kg every 3 weeks and 13 were randomized to receive durvalumab 10mg/kg every 3 weeks. Grade 3/4 TRAEs were reported in 68.8% of dose-escalation patients, leading to discontinuation of combined treatment in 94% of patients in this phase. In the dose-expansion phase (Part B), 40 patients were recruited to one of two treatment strategies: 1) gefitinib + durvalumab 10 mg/kg every 2 weeks (Arm 1, 30 patients), or 2) gefitinib for 4 weeks followed by addition of durvalumab 10 mg/kg every 2 weeks (Arm 2, 10 patients). Median PFS was 10.1 months in Arm 1 (95% CI 5.5-15.2 months) and 12.0 months in Arm2 (95% CI: 2.7-15.6 months), which was not considered improved compared to gefitinib monotherapy studies. For example, the phase 4 update of the NCT01203917 first-line gefitinib study demonstrated PFS of 9.7-10.2 months with gefitinib monotherapy (59). 17/40 of the dose-expansion phase patients experienced high-grade hepatic events, suggesting an additive effect of gefitinib and durvalumab for hepatotoxicity. The authors noted a trend towards favorable PFS in patients with TPS ≥ 20% (HR 0.46, 95% CI 0.19-1.03), but overall concluded that their study did not support combination gefitinib + durvalumab as first line treatment in EGFRm advanced NSCLC.

The ongoing open-label, multicohort phase 1/2 KEYNOTE-021 study (NCT02039674) is evaluating pembrolizumab in combination with chemotherapy or immunotherapy in previously untreated stage IIIB/IV EGFRm NSCLC patients (41). Two cohorts were reported early and evaluated pembrolizumab + EGFR TKI: Cohort E included 12 patients treated with pembrolizumab + erlotinib and Cohort F included 7 patients treated with pembrolizumab + gefitinib. The pembrolizumab + gefitinib arm was permanently discontinued early due to safety concerns, as 5/7 patients (71.4%) had treatment-related elevations in ALT and AST. However, the pembrolizumab + erlotinib arm was found to be tolerated and this arm experienced an objective response rate (ORR) of 41.7%, which is similar to the ORR seen with erlotinib and pembrolizumab monotherapies. All (4/4) patients with TPS ≥ 50% had an objective response, whereas only 1/4 patients with TPS 1%-49% and 0/2 patients with TPS <1% responded to pembrolizumab + erlotinib (41).

The phase 1b NCT02013219 study evaluated first-line erlotinib + atezolizumab in TKI-naïve patients with EGFRm NSCLC (42). The preliminary report included 28 patients and demonstrated a median PFS of 11.3 months, which was similar to erlotinib monotherapy. However, 50% of patients experienced serious TRAEs including 39% of patients who experienced grade 3/4 events. These data suggest that combination atezolizumab + erlotinib enhances toxicity without significant additive benefit. We await the publication of full trial results make a more comprehensive assessment of this combination.

The Phase 1/2 CheckMate 370 study (NCT02574078) is evaluating nivolumab as maintenance or first-line + other standard of care therapies (60). Group D will compare erlotinib versus nivolumab + erlotinib. Results have not yet been announced.

As part of the multi-arm phase 1 CheckMate 012 (NCT01454102) trial, patients with chemotherapy-naïve Stage IIIB/IV NSCLC were randomized to receive different dose schedules of nivolumab and ipilimumab as first line treatment. 10-11% of patients in different arms had EGFR activating mutations. Of these, 50% (4/8) had objective responses with combined nivolumab + ipilimumab, including 3/3 (100%) of patients with TPS ≥ 50% (43). PFS data are not available. These results, while limited by small sample size, suggest that combination immune checkpoint inhibition therapy may more effectively sensitize EGFR-mutant, PD-L1-expressing NSCLC to immune-mediated destruction than ICI monotherapy.

In regard to single agent ICI as a second-line agent after PD on EGFR TKI, three studies explored different single agent ICIs versus docetaxel. KEYNOTE-010 included a small number of EGFRm advanced NSCLC patients (9%) who were randomized to receive 2mg/kg pembrolizumab, 10mg/kg pembrolizumab, or 75 mg/m² docetaxel every three weeks after having progressed on at least two cycles of platinum-based chemotherapy and treatment with an EGFR TKI. The EGFRm patients experienced worse PFS with pembrolizumab versus docetaxel (HR for PFS 1.79, 95% CI 0.94-3.42, p not given), in contrast to the EGFRwt patients who had an improved pooled PFS (HR for PFS 0.83, 95% CI 0.71-0.98, p not given) (44). CheckMate 057 (NCT01673867) was a large phase 3 study of nonsquamous NSCLC patients who had progressed on or after platinum-doublet therapy that compared nivolumab 3mg/kg every two weeks versus docetaxel 75 mg/m² every three weeks. EGFRm patients were allowed to have received previous treatment with an EGFR TKI. 14% of study participants were EGFRm positive and experienced worse OS with nivolumab compared to docetaxel (HR 1.18, 95% CI 0.69-2.00, p not given), whereas the EGFRwt patients experienced significant benefit to OS with nivolumab (HR 0.66, 95% CI 0.51-0.86, p not given) (45). The POPLAR phase 2 study compared atezolizumab versus docetaxel as second line therapy for NSCLC. Similar to the PD-1 inhibitor studies, atezolizumab improved OS compared to docetaxel among all NSCLC patients (HR 0.73, 95% CI 0.53-0.99, p=0.04) (46). However, as reported by Lee et al., the subgroup of EGFRm patients did not see an improvement in OS (HR for OS 0.70 in WT versus 0.99 in EGFR-mutant, compared to docetaxel) (47). Thus, PD-L1 inhibition, as with PD-1 inhibition, failed to demonstrate improvement in EGFRm patients in the second-line context, though ICI therapy was more well tolerated across all three studies with less TRAEs compared to docetaxel.

A meta-analysis published by Lee et al. combined the results of these three studies utilizing ICI single therapy versus docetaxel as second-line therapy. The EGFRm NSCLC patients overall did not benefit from ICI monotherapy compared to docetaxel, whereas EGFRwt patients experienced a significant benefit from ICI therapy (HR for OS 1.05 for EGFRm vs 0.66 for EGFRwt) (47), confirming that ICI monotherapy is not advantageous over chemotherapy in the second-line setting for EGFRm patients with PD on TKI/chemotherapy.

Hui et al. reported updated results from the KEYNOTE-001 study, which examined pembrolizumab efficacy across several settings for patients with NSCLC and PD-L1 TPS ≥ 1% (48). The subgroup of EGFRm patients who had previously received treatment had significantly less benefit from pembrolizumab than EGFRwt patients (median OS 12.1 months versus 6.0 months). PD-L1 overexpression (TPS ≥ 50%) did not rescue response to pembrolizumab in the EGFRm vs EGFRwt patients (median OS 6.5 versus 15.7 months) (48). These results suggested that PD-L1 is an imperfect biomarker to predict ICI response in previously treated EGFRm patients, as discussed in further detail below.

The PACIFIC trial (NCT02125461) was a phase 3 trial that assessed the addition of durvalumab consolidation therapy after definitive chemoradiotherapy (CRT) for patients with stage III NSCLC (49). EGFRm NSCLC patients did not have significant benefit from durvalumab consolidation therapy (HR for PD or death 0.76, 95% CI: 0.35-1.64) whereas the EGFRwt patients did experience benefit (HR for PD or death 0.47, 95% CI: 0.36-0.60). In the recently published four-year survival update of the PACIFIC trial, EGFRm NSCLC patients again did not demonstrate benefit from durvalumab consolidation (HR for PFS 0.84, 95% CI: 0.40-1.75) whereas the EGFRwt patients again demonstrated significant benefit (HR for PFS 0.51, 95% CI: 0.40-0.65) (61).

Similar to the PACIFIC trial, Aredo et al. recently published the results of a multi-center retrospective study of patients (n=13) with unresectable EGFRm NSCLC who received consolidation durvalumab after CRT (62). They compared these patients to a cohort of EGFRm NSCLC patients who instead received consolidation EGFR TKI after CRT (n=24). Median PFS was 10.3 months for the CRT + durvalumab cohort versus 26.1 months for the CRT + EKGFR TKI group (p = 0.023). Notably, six patients opted to switch to EGFR TKI after experiencing PD on CRT + durvalumab and one of these patients developed Grade 4 pneumonitis 17 days after initiating osimertinib, again highlighting the safety signal of initiating EGFR TKIs after ICI therapy.

The ATLANTIC trial was a phase 2 open-label trial of durvalumab monotherapy as third-line or later treatment in patients with advanced NSCLC (63). Enrolled patients had to have received at least two previous lines, with one platinum-containing regimen and a TKI if indicated. Cohort 1 included both EGFRm and ALKm NSCLC patients and was stratified by PD-L1 TPS: median OS was 13.3 months in the TPS ≥ 25% subcohort versus 9.9 months in the TPS < 25% subcohort. Notably, this was higher than in the EGFRwt and ALKwt Cohort 2 that demonstrated median OS of 10.9 versus 9.3 months for TPS ≥ 25% and TPS <25% subcohorts, respectively. The safety profile of EGFRm/ALKm cohort was similar to the EGFRwt/ALKwt profile, with 6-8% of patients experiencing Grade 3-4 TRAEs, supporting the safety of ICI administration after EGFR TKI treatment. Of note, the ATLANTIC trial has multiple limitations, including lack of descriptive statistics, single-arm design, and variations in testing platform for PD-L1 that made direct comparisons with other trials not possible (64). Despite this, the ATLANTIC trial suggested that ICI therapy may have efficacy in heavily pre-treated EGFRm NSCLC patients and further supported the safety of ICI therapy after patients progressed on TKI therapy.

The NCT03513666 trial was a phase 2 study that evaluated toripalimab (anti-PD-1) + platinum doublet chemotherapy in patients with EGFRm advanced NSCLC who developed PD on first- and second-generation EGFR TKIs without T790M mutation (51). Median PFS was 7.0 months, and interestingly the authors identified that TP53 co-mutation patients experienced significantly improved ORR compared to TP53wt patients (62% vs 14%, p = 0.04). This combination was found to have manageable safety profile and efficacy, and a follow up randomized Phase III trial (NCT03924050) that will compare this combination to standard chemotherapy with planned enrollment for 350 patients (65).

The CheckMate 722 trial (NCT02864251) is a currently active phase 3 study of patients with EGFRm, T790M-negative recurrent or stage IV NSCLC who have previously been treated with EGFR TKI therapy (66). Arm A comprises nivolumab + platinum-doublet therapy and Arm C involves platinum-doublet alone. Additionally, the KEYNOTE-789 trial (NCT03515837) is another currently active Phase III trial evaluating pemetrexed-platinum combined with pembrolizumab versus placebo in EGFRm advanced NSCLC that has progressed on EGFR TKI (67). The results of these studies will provide highly valuable information on the efficacy and safety of second line ICI + chemotherapy. The CheckMate 722 trial has the additional benefit of comparing this strategy to dual ICI therapy (Arms A versus B).

A number of trials have evaluated combination ICI therapy and EGFR-targeted therapy in the second-line and beyond. These studies were predicated on pre-clinical studies that suggested added benefit to the combined approach in animal models (68–71). However, results have mostly been disappointing and in many cases demonstrated increased and severe toxicities. Nonetheless, several ongoing trials are assessing different combination therapies that will be of interest when results are available.

Cohort H of the KEYNOTE-021 phase 1/2 study assessed pembrolizumab 2 mg/kg plus ipilimumab 1 mg/kg as second-line or later therapy (55). Of the 10 EGFRm patients, only 1 (10%) patient responded to therapy compared to an ORR of 30% for the entire study population. 98% of patients experienced a TRAE, including 49% with grade 3-5 AEs.

Multiple active studies are investigating dual ICI therapy in second line or later EGFRm NSCLC. The ILLUMINATE Phase 2 study (NCT03994393) is evaluating the safety and tolerability of combined durvalumab and tremelimumab plus platinum-pemetrexed in EGFRm NSCLC following progression on EGFR TKIs (75). 100 patients will receive induction durvalumab + tremelimumab with platinum-pemetrexed every three weeks, followed by maintenance durvalumab + pemetrexed every four weeks until disease progression. EGFRm T790M negative and positive patients will be included. Additionally, two arms of the CheckMate 722 phase 3 trial, described above, will compare nivolumab + ipilimumab (Arm B) to platinum-doublet chemotherapy (Arm C). We eagerly await the results from these study that will leverage the potentially enhanced immune response of dual ICI therapy.

The IMpower150 Phase 3 study (NCT02366143) assessed the addition of PD-L1 inhibition with atezolizumab and VEGF inhibition with bevacizumab to carboplatin + paclitaxel (CP) in patients with mNSCLC. Regimens included bevacizumab + CP (BCP), atezolizumab + CP (ACP), and atezolizumab + BCP (ABCP) (76). A subgroup analysis was performed of EGFRm patients; notably, 85-88% of the patients had previously received at least one EGFR TKI therapy (56). In the initial subgroup analysis, median OS was not reached with the ABCP group in EGFRm patients. Fortunately, the updated results were recently published and demonstrated that among EGFRm patients who had received prior TKI therapy a significant increase in median OS was observed with the ABCP regimen (27.8 months versus 14.9 months with ACP and 18.1 months with BCP) (77). The HR for ABCP versus BCP was 0.74. The toxicity profile was similar between different regimens in EGFRm patients: 64-68% of patients across all three regimens experienced at least one Grade 3-4 TRAE, and 1-3% experienced a Grade 5 TRAE. These results suggested the hopeful possibility of an effective ICI-combination therapy for patients who experienced PD on an EGFR TKI.

Lam et al. recently reported the results of the Phase II NCT03647956 trial that enrolled 40 patients with metastatic EGFRm NSCLC that had progressed on EGFR TKI (57.5% osimertinib) (57). Patients were treated with atezolizumab + bevacizumab + pemetrexed-carboplatin until progression. Median PFS was 9.43 months and median OS was not mature yet at time of publication (1-year OS was 72.5%). 37.5% of patients experienced a grade 3 or above TRAE but only 1/40 patients discontinued treatment due to toxicity. These encouraging results, coupled with the IMpower150 EGFRm subgroup results, support the potential efficacy of adding VEGF inhibition to ICI and chemotherapy as second-line in EGFRm NSCLC that has progressed on EGFR TKI.

The phase 2 NCT04517526 trial is planning to enroll 60 patients with stage IV EGFRm NSCLC with PD after first-line osimertinib. Patients will receive platinum-based chemotherapy + bevacizumab + durvalumab + stereotactic radiotherapy to oligometastatic or oligoprogressive sites (78). The results of this study will be of great interest as it will combine advanced combination immunotherapy and radiation therapy approaches.

PD-L1 TPS is a standard biomarker for stratifying patients in clinical trials of ICI therapy in EGFRwt cells, with evidence from multiple trials that higher PD-L1 TPS tumors (e.g. with TPS ≥ 50%) have an enhanced response to ICI therapy (84, 85). Of note, PD-L1 expression is not uniformly prognostic of response to ICI therapy (86), and it has been demonstrated that NSCLC patients without PD-L1 immunohistochemical staining can still derive benefit from ICI therapy (87), supporting the now widely accepted notion that PD-L1 status by itself is insufficient to predict ICI response.

In EGFRm NSCLC, the value of PD-L1 expression is even less clear. Mechanistic studies have demonstrated that upregulation of EGFR signaling in vitro leads to increased PD-L1 expression by pathways including the IL-6/JAK/STAT3 pathway (88). However, immunohistochemical and mRNA expression profiling analysis of EGFRm NSCLC patient tumor samples demonstrated decreased PD-L1 expression across multiple datasets (89, 90), leading to an unresolved discrepancy between preclinical and clinical studies. Of note, a weakness of these studies is the inability to assess the half-life of PD-L1 between subgroups.

In regard to treatment response, multiple studies have demonstrated that increased PD-L1 expression on EGFRm NSCLC cells predicts worse outcomes with TKIs but improved outcome with ICI therapy (91–93). Liu et al. recently published a correlation analysis of 57 EGFRm NSCLC patients who received ICI treatment after developing PD on EGFR TKIs (94).They identified by using a TKI-PFS cutoff of 10 months that EGFRm patients with <10 month TKI-PFS had significantly improved ICI-PFS of 15.1 versus 3.8 months, respectively (HR 0.26, 95% CI: 0.12-0.5, p = 0.0002), strongly suggesting that EGFRm tumors are either TKI or ICI responsive. Intriguingly, this relationship was independent of PD-L1 status, again reiterating the importance of other elements of the TME in treatment response. To further probe differences in the TME between these groups, they performed single-cell RNA sequencing of patients with TKI-PFS <10 months (group A) and >10 months (group B). Group A demonstrated significantly higher proportion of T-cell TILs along with increased CD8+ effector proportion of T-cells. These results suggest a critical role for CD8+ effector TILs in determining response to ICI therapy in EGFRm NSCLC, as discussed in more detail below.

While PD-L1 expression is not uniformly predictive of ICI response, there is some evidence that EGFRm NSCLC tumors with increased PD-L1 TPS have improved response to third or later line ICI therapy (63). Thus, while ICI monotherapy is inappropriate for first line treatment as described above, PD-L1 analysis may be a valuable component of a holistic evaluation of the TME in EGFRm, along with other elements including TMB, TILs, and other discussed below to assist oncologists in deciding on later-line ICI treatment strategies for EGFRm patients who fail EGFR TKI therapy. As such, future clinical trials should continue to gather and report PD-L1 expression data so that these relationships can be better elucidated.

The efficacy of ICI therapy depends on the intratumoral migration and activation of CD8⁺ effector T-cells where they perform cytotoxic functions after interaction of the T-cell receptor with tumor-specific peptides displayed on MHC-I complexes on tumor cells (29). Multiple studies have demonstrated that EGFRm NSCLC tumors have reduced CD8⁺ TIL presence compared to EGFRwt tumors (89, 95, 96). Interestingly, Zhao et al. recently published an analysis of 190 surgical lung ADC samples that demonstrated increased apoptosis in the EGFRm patient tumor samples (96). They further went on to demonstrate that exosomes secreted from EGFRm cells were more capable of inducing CD8⁺ T-cell apoptosis in vitro than exosomes from EGFRwt cells. These results suggest that, in addition to reduced TIL density in EGFRm tumors, there may also be increased TIL apoptosis that impairs immune-mediated tumor destruction. Further study of this mechanism may provide valuable new information on the TME in EGFRm NSCLC and possibly provide a novel therapeutic avenue to enhance antitumor immunotherapy (97).

Strategies to increase TIL trafficking and activity in tumors including NSCLC are an active area of research (98). Possible approaches include targeted tumor necrosis factor α (TNF-α) delivery and anti-angiogenic drugs including inhibitors of VEGF. TNF-α causes endothelial cell activation and increased vessel permeability that can enhance the ability of both chemotherapy and immune cells to penetrate solid tumors (99); however, systemic TNF-α administration is quite toxic (100). A compound containing the tumor vasculature-homing peptide Cys-Asn-Gly-Arg-Cys (NGR) has been fused to TNFα to create a tumor vasculature-homing version of TNF-α that avoids the toxicity of systemic TNFα administration (NGR-hTNF) (101). A phase II trial of NGR-hTNF combined with chemotherapy in patients with chemotherapy-naïve NSCLC was previously reported. Patients with nonsquamous NSCLC in the chemotherapy + NGR-hTNF arm experienced improved PFS at 8 months (38% versus 18% with chemotherapy alone) and a tolerable safety profile (102). While preclinical models support the ability of TNF-α to also enhance ICI therapy (103), there are no active clinical trials involving NGR-hTNF and ICI therapy in NSCLC due to the manufacturer of NGR-hTNF discontinuing the product after a Phase III mesothelioma trial did not meet its primary endpoint (104). Importantly, an NGR-TNF derivative with an additional serine at the N-terminus that demonstrates increased stability, S-NGR-TNF, has been recently developed (105). It will be intriguing to see if TNF-α strategies such as S-NGR-TNF can restore TIL trafficking, enhance ICI therapy, and augment chemotherapy delivery to EGFRm NSCLC tumors.

VEGF is also known to suppress TILs via multiple mechanisms, including suppressing endothelial cell activation, inhibiting TNFα-mediated gene regulation, and blocking dendritic cell maturation thereby reducing T-cell activation (106, 107). In preclinical models, VEGF inhibition synergized with PD-1 blockade and reduced T-cell exhaustion, and in clinical studies combinations of therapies including ICIs and TKIs have demonstrated improved TIL recruitment and improved PFS (108). The IMpower150 study, for example, demonstrated improved PFS of EGFRm NSCLC patients who had disease progression on or did not tolerate an EGFR TKI when they were treated with a combination of chemotherapy, ICI therapy, and VEGF-inhibition as described above (56, 76). The success of VEGF-inhibition in second-line combination chemotherapy + ICI therapy for EGFRm NSCLC patients with PD on EGFR TKI is one of the few bright signals currently in the field, and we eagerly await the results of current trials that are ongoing further exploring this question ( Table 3 ).

On average, EGFRm NSCLC patients have a decreased TMB compared to EGFRwt patients (89, 109). This is thought to be at least partly due to the fact that EGFRm patients tend to have a lighter smoking history. Increased TMB classically correlates to decreased response to chemotherapy and an increased response to ICI therapy in NSCLC (110, 111), and in EGFRm NSCLC patients increased TMB correlates negatively with response to EGFR TKIs (112). Increased TMB is thought to potentiate ICI therapy by creating an environment where more tumor-specific neoantigens are generated, thus creating more targets for TILs to recognize and enhancing the adaptive anti-tumoral response. Indeed, significantly fewer candidate MHC class-I neoantigens were identified in EGFRm versus EGFRwt NSCLC tumors in a whole-genome DNA sequencing study (113). While the decreased TMB in EGFRm NSCLC patients overall may contribute to decreased efficacy of ICI therapy, there is some evidence that TMB may still be of significance in this population. For example, Hastings et al. retrospectively analyzed 171 cases of EGFRm NSCLC and demonstrated that EGFR^Δ19 tumors had a lower TMB and a worse response to ICI therapy compared to EGFR ^L858R tumors (114). Additionally, certain hypermutator phenotypes such as DNA mismatch repair (MMR) deficient tumors and DNA polymerase delta and epsilon proofreading mutants, while uncommon in NSCLC, may respond well to ICIs (115, 116). As such, subgroups of EGFRm patients with increased TMB, while less common than in EGFRwt context, are predicted to still receive increased benefit from ICI therapy compared to their low TMB counterparts. TMB analysis is an intriguing and significant element that should be strongly considered for clinical studies of EGFRm NSCLC patients.

CD73 is an ecto-nucleotidase that catabolizes the breakdown of extracellular ATP to adenosine (117). There is a growing appreciation that the CD73/adenosine axis plays a significant and complex role in the TME. Increased intratumoral adenosine contributes to localized immunosuppression and impairment of T-cell effector function (118, 119), and the CD73/adenosine axis is becoming considered an immune checkpoint in its own right (120). Pre-clinical data demonstrated that anti-CD73 monoclonal antibodies (mAbs) significantly enhanced the activity of anti-CTLA-4 and anti-PD-1 mAbs in animal studies of colon, prostate, and breast cancer (118). An intriguing, recently published study by Le et al. analyzed upregulated genes in EGFRm NSCLC tumors and found that two of the top upregulated genes (NT5E and ADORA1) belonged to the CD73/adenosine pathway (89), suggesting that EGFRm NSCLC may leverage the CD73/adenosine axis to generate an immunosuppressive TME. They assessed the efficacy of an anti-CD73 mAb in a mouse model of EGFRm murine lung cancer and found that anti-CD73 treatment significantly reduced tumor size. As such, an active question is whether suppression of the CD73/adenosine axis can enhance the treatment of EGFRm NSCLC. Along these lines, a human mAb targeting CD73, Oleclumab, is being assessed in a phase 1b/2 study (NCT03381274) in combination with either osimertinib or AZD4635, which is an adenose 2a receptor (A2aR) inhibitor (121). Given the encouraging preclinical data, we eagerly look forward to further clinical trials utilizing anti-CD73 mABs or A2aR inhibitors in conjunction with ICI therapy.

There is intriguing evidence that the specific EGFR mutation impacts the immunogenicity of the TME and response to ICI therapy. Chen et al. performed a large single-study of 600 NSCLC patients in China with EGFRm NSCLC and identified 49 with uncommon mutations (Ex20ins, S767I, L861Q, G719X, and double mutations) (91). They found a much higher proportion of PD-L1 expressing tumors with uncommon mutations compared to classic mutations (49% versus 12.2%), and CD8⁺ TIL infiltration was more abundant in this group (91). They reported worse OS for patients with PD-L1 positive EGFRm NSCLC versus PD-L1 negative (median OS 15.2 versus 29.3 months, p = 0.006), though most of these patients received EGFR TKI monotherapy across all lines of treatment. Negrao et al. reported that metastatic EGFRm exon 20 mutation NSCLC patients had increased benefit from ICIs compared to classic mutation patients (exon 19 del, exon 21 L858R) with an ORR of 25% versus 0% and disease control rate (DCR) of 50% versus 15% (122). Mazieres et al. analyzed the IMMUNOTARGET registry and compared the molecular characteristics of EGFRm patients to response to ICIs (123). EGFR exon21 mutation patients derived significantly longer PFS from single-agent ICI therapy (2.5 months) in this database than patients with T790M and exon 19 mutations (1.4 and 1.8 months, respectively, p <0.001). As such, these studies suggest that uncommon EGFR mutations including exon 21 mutations may have increased immunogenicity and response to ICIs. Future clinical trials should ensure that the specific EGFR genetic alterations are reported and provide mutation subgroup data so that further evidence on this subject can be obtained.

One of the greatest needs in the field currently is the development of a scoring/stratification system that will predict which EGFRm patients will benefit from ICI therapy. As more studies publish the results of detailed molecular and immunohistochemical analysis, this will empower a more comprehensive understanding of the cellular composition of EGFRm TMEs and tumor biologies ( Figure 1 ). Complex multivariate analyses should be employed to delineate subgroups of EGFRm patients that will benefit from ICI therapy. The use of artificial intelligence (AI) including artificial neural networking is being studied for the analysis of TMEs (124, 125), and may prove invaluable to identify signatures of EGFRm tumors that predict ICI response. The value of machine learning in EGFRm tumor biology was recently demonstrated by Song et al. who utilized a machine learning model to analyze pre-treatment computed tomography (CT) images of stage IV EGFRm NSCLC patients (126). Their machine learning approach successfully identified an imaging signature able to stratify EGFRm patients most likely to rapidly progress despite TKI therapy. It is a logical next step to apply machine learning to stratify patients likely to respond to ICI therapy based on tumor biological characteristics.

Multiple lines of pre-clinical evidence suggested synergy between EGFR TKI inhibition and ICI therapy. In pre-clinical studies, EGFR inhibition enhanced antigen presentation to T-cells, stimulated immunogenic apoptosis of tumor cells, boosted T-cell chemoattractants, and stimulated MHC-1 upregulation, all of which are predicted to enhance the anti-tumor immune response (68–71). Despite this, early clinical studies demonstrated that preclinical studies would not translate in a straightforward manner. IHC analysis of tumors from early EGFRm patients treated with ICIs, against expectation, demonstrated decreased PD-L1 expression and decreased CD8+ TILs (95), data that has since been recapitulated in multiple studies described above. This has led to the active research question of the effect of EGFR TKIs on the TME of EGFRm NSCLC in vivo during and after therapy.

Multiple groups have addressed this question with TME analysis at various time points of treatment. Isomoto et al. performed serial immunohistochemical analysis of 138 patients who underwent rebiopsy after progression on EGFR TKI treatment (127). They found multiple significant changes in the TME after PD, including an expanded proportion of high (≥50%) PD-L1 expressing tumors and decreased CD8+ TILs in PD-L1 <50% tumors. Notably, they identified subgroups with opposing clinical courses: tumors with high PD-L1 expression after progressing on EGFR TKI had significantly longer PFS with ICI therapy (7.1 versus 1.7 months, p = 0.0033) and increased CD8⁺ TIL presence. In contrast, PD-L1 <50% tumors had significantly decreased CD8⁺ TIL density. Also of interest, the PD-L1 high tumors had increased FOXP3+ and CD73 TIL density, suggesting that regulatory T-cell (Treg) and CD73 axis activation may contribute to ICI treatment failure.

Sugiyama et al. analyzed surgically resected EGFRm tumors and found decreased CD8+ TILs and increased FOXP3+CD4+ Tregs, further supporting a role for Treg suppression of the immune response in EGFRm tumors (128). Gurule et al. performed RNA sequencing of patient tumors before and 2 weeks after TKI treatment and demonstrated induction of an interferon response program (71). Interestingly, higher enrichment of interferon gamma (IFNγ) was correlated with longer time to progression. Taken together, these results suggest that EGFRm NSCLC tumors undergo diverse responses to EGFR TKIs with some tumors becoming more immunogenic and some becoming more immunosuppressive with resulting divergent responses to ICI therapy (129). While the mechanism behind the divergent TME responses to TKI therapy in EGFRm NSCLC is unclear, these studies suggest that rebiopsy may have clinical benefit in identifying subpopulations of patients who are more likely to respond to ICI as subsequent therapy.