Histo-Molecular Factors of Response to Combined Chemotherapy and Immunotherapy in Non-Small Cell Lung Cancers

Over the past 15 years, treatment of advanced non-small cell lung cancer (NSCLC) has undergone profound changes, with the development of targeted therapies and immune checkpoint inhibitors (ICIs). These therapeutics offer the possibility of long-term disease control. Meanwhile, histo-molecular profiling has become part of routine care, its aim being to detect targetable oncogenic alterations and biomarkers predictive of response to both targeted therapies and immunotherapy [1].

Inhibitors of programmed cell death-1 (PD1) or programmed cell death ligand-1 (PDL1) have significantly improved the prognosis of patients with advanced NSCLC, but their efficacy is in fact variable, and only a subset of patients actually derives a prolonged benefit from immunotherapy [2, 3]. Several potential biomarkers, reflecting the balance between the pro-inflammatory anti-tumor immune response and immune evasion mechanisms, may allow more accurate prediction of the clinical response to immunotherapy, but only the PDL1 tumor proportion score (TPS) is currently used in clinical practice [4].

Testing for a driver oncogene alteration may help determine which patients may benefit from ICI treatment as monotherapy. Anti-PD(L)1 agents are associated with poor outcomes in patients with EGFR mutations and ALK rearrangements [5‐8]. This lack of sensitivity may be related to an uninflamed tumor micro-environment and a low tumor mutational burden (TMB) [9, 10]. As a consequence, patients with EGFR mutations or ALK rearrangements have been excluded from most clinical trials evaluating anti-PD(L)1 as first-line therapy. In contrast, KRAS mutations appear to be predictive of a good response to ICI treatment, which can be explained by their association with an inflammatory tumor microenvironment and a high PDL1 TPS, a high CD8+ T-cell infiltration, and a high TMB, which may reflect the strong epidemiological link with smoking [11, 12]. However, interactions between anti-PD(L)1 agents, tumor cells, and host immunity are complex, and co-mutations may impact the efficacy of ICI treatment. For example, STK11 mutations correlate with a poor response rate [13].

Other predictive markers of the efficacy of ICIs have been gradually explored from the tumor immune microenvironment (phenotype of tumor infiltrating lymphocytes, T-cell receptor repertoire diversity) to the exploration of circulating and host systemic markers (peripheral blood cells, inflammatory circulating proteins, Human Leukocyte Antigen-I variability, intestinal commensal microbiota) but are not yet integrated into clinical practice [14].

It has been demonstrated that adding anti-PD(L)1 to standard chemotherapy increases the chances of response and prolongs survival [12, 15]. Consequently, the combination of chemotherapy with immunotherapy (chemo-immunotherapy, CIT) has become a standard of care for advanced NSCLC [1]. The benefits observed with this combination may result from synergy between the immune response induced by ICIs and the immunogenic effects of chemotherapy, such as an increased release of tumor antigens (immune cell death), inhibition of myeloid-derived suppressor cells, or an increased ratio of cytotoxic lymphocytes to regulatory T cells [16]. Given the specificities of the mechanism of action of CIT, the usual biomarkers of response to PD(L)1 pathway inhibition might not be relevant to CIT. In particular, it remains to be seen whether the poor sensitivity to immunotherapy observed in patients with oncogene addiction holds true when chemotherapy is added. On the basis of the anti-tumor immunological effects of chemotherapy, one can hypothesize that adding chemotherapy to immunotherapy might promote recovery of immunosurveillance function and restore the activity of ICIs in ‘cold’ tumors. We thus conducted a multicenter observational study to describe the efficacy of CIT according to histo-molecular factors.

In our real-world study, PFS was lower than in the KEYNOTE-189 clinical trial (9.0 months) and also lower than in a retrospective study evaluating the real-word effectiveness of CIT in untreated patients (8.6 months) [15, 23]. These results might be due to the higher proportion of pre-treated patients and patients with impaired PS included in our study.

However, it does not appear to be related to the inclusion of patients with EGFR mutations or ALK rearrangements. Indeed, although the lack of statistical difference in PFS between driver subgroups may be due to the lack of power because of the low number of patients in the EGFR/ALK/ROS1/RET  and BRAF subgroups, response to CIT did not appear to be impaired in patients with EGFR mutations and rearrangements in our study.

The place of CIT in patients with EGFR mutations or key rearrangements, after progression on tyrosine kinase inhibitor, is still under discussion. Whereas EGFR mutations and ALK rearrangements are associated with a lack of efficacy with immunotherapy as monotherapy [10, 24], our study did not show poor outcomes in the EGFR and rearrangements subgroups. These results are consistent with the fact that smoking was not associated with PFS in our study. However, we cannot assess whether the benefit is due to the addition of immunotherapy, synergy with the immunological effects of chemotherapy, or chemotherapy alone. Pemetrexed has indeed been shown to be associated with a favorable outcome in oncogene-addicted cancers, especially ALK-rearranged NSCLC [25]. Shen et al. and Benjamin et al. reported PFS durations close to those of our study in patients with EGFR mutations treated by CIT [26, 27].

The efficacy of CIT in NSCLC patients with EGFR mutations is uncertain. The ORIENT-31 clinical trial reported a PFS benefit with sintilimab plus chemotherapy compared with chemotherapy alone, while KEYNOTE-789 failed to demonstrate a significant improvement of PFS and OS with the addition of pembrolizumab to chemotherapy [28, 29]. The addition of bevacizumab appears to have a potentially beneficial role in some cases [29‐31]. However, further research is needed to determine the best treatment for these patients. No patients treated with CIT plus bevacizumab were identified in our study.

Despite close clinical and pathophysiological characteristics and similar profiles of response to immunotherapy as monotherapy [7, 8, 32], HER2-mutated patients seemed, in our study, to have derived less benefit from CIT than those with EGFR mutations. Yang et al. reported that CIT did not improve PFS compared with chemotherapy alone in a retrospective cohort of 210 HER2-mutated patients (mPFS: 5.2 months vs 4.0 months, HR 0.77 [95% CI 0.52–1.14]) [33]. Saalfeld et al. showed better outcomes with CIT (mPFS: 6 months; ORR: 52%) than with ICI alone (mPFS: 4 months; ORR: 16%) in 61 HER2-mutated patients [34]. We report poorer results in these patients, but the number of patients was very small, which may have led to sampling variance. Additional studies are needed to better assess the impact of HER2 mutations on the efficacy of CIT.

In our study, patients with BRAF mutations derived the greatest benefit from CIT. These results are mainly driven by BRAF V600E patients, although most of these were never-smokers (67%) and only one had high expression of PDL1 ≥ 50%. A case report has also shown a prolonged response to CIT in a never-smoker BRAF V600E NSCLC patient [35]. Retrospective studies having investigated the efficacy of immunotherapy alone in patients with BRAF mutations have reported a mPFS in the range of 1.5–5.3 months [36]. These results, if confirmed, would raise the question of the best therapeutic sequence between CIT and targeted therapies, as already discussed for BRAF-mutated melanoma patients.

Patients with MET exon 14 mutations or MET amplification showed the poorest outcomes. Median PFS was similar in the two patients with MET amplification (1.5 and 1.6 months) but varied markedly among the four patients with MET exon 14 skipping, from 0.6 to 8.6 months. Biological studies suggest that the MET/HGF axis plays an immunomodulatory role in the tumor microenvironment [37, 38], but favorable outcomes have been reported in several studies evaluating ICI as monotherapy in the case of MET alterations [36]. Given these discrepant results and the small number of patients with MET alterations included, further studies are needed to specify the role of these alterations in responses to CIT. However, the advanced age of patients with MET alterations may partly explain these poor results [39].

Two meta-analyses have shown that KRAS mutations are a potential biomarker of a survival benefit from ICI treatment, because they promote a proinflammatory tumor micro-environment [40, 41]. In our study, PFS did not differ statistically between the KRAS subgroup and other patients, but the Kaplan-Meir curves seem to be in favor of the KRAS patients. These results are consistent with an exploratory analysis of the KEYNOTE-189 trial [42]. In the IMPOWER-150 trial, however, survival with ACP (atezolizumab plus carboplatin plus paclitaxel) was shorter in KRAS-mutated patients compared with patients without KRAS mutation [43]. On the basis of these findings, the role of KRAS mutations as predictive biomarkers of response is less clear with CIT than with immunotherapy alone.

STK11 mutations appeared to be a factor of poor prognosis in the whole cohort and particularly in the KRAS subgroup. They have likewise been shown previously to be associated with poor prognosis in patients receiving ICI treatment [43, 44]. Although strongly associated with smoking, STK11 inactivation has been shown to result in a cold immunosuppressive environment characterized by low PDL1 TPS and a low density of T CD8+ lymphocytes [44]. Accordingly, among the STK11-mutated patients included in our study, 51% had PDL1 TPS < 1% and only one had PDL1 TPS ≥ 50%. Yet, low PDL1 expression is insufficient to explain the fact that STK11-mutated patients do not benefit from the synergistic immunological effects of CIT. Further investigations are needed to understand the pathways underlying impairment of the antitumor immune response in the case of STK11 mutations, but several potentially interrelated mechanisms have already been outlined: downregulation of chemokines and interferons α and β, increased immunosuppressive cytokines, metabolic competition with immune effector cells, enhanced prostaglandin production, and impairment of dendritic cell recruitment [13].

Another tumor suppressor, p53, leads to accumulated DNA damage and thus to increased tumor immunogenicity. In two reported series, TP53 mutations were found to correlate positively with ICI efficacy [45, 46]. In the IMPOWER-150 trial, TP53 mutations were associated with better outcomes with ABCP (atezolizumab plus bevacizumab plus carboplatin plus paclitaxel), but not with ACP, in KRAS patients [43]. In our study, no difference in efficacy was observed according to the presence of TP53 mutations, but the PFS Kaplan-Meier curves tended to separate at 12 months and beyond in favor of TP53-mutated patients.

MET overexpression has been linked to poor prognosis in NSCLC patients, in line with the biological properties of MET, known to promote proliferation and invasiveness [21, 47]. Surprisingly, in our study, MET overexpression was associated with better efficacy and longer survival, although the amount of missing data limits the scope of the results. To our knowledge, this is the first description of the efficacy of anti-PD(L)1-based therapy according to the level of MET overexpression. MET expression has been found to be associated with high PDL1 expression and tumor-infiltrating lymphocytes in NSCLC [48]. The reason MET overexpression correlated with survival while patients harboring MET exon 14 mutations and MET amplification showed poor outcomes is unclear. Some points worth considering could be the lack of correlation between MET overexpression and MET inhibitor efficacy and differences in patient characteristics, especially in terms of age and smoking status [49‐51].

The small number of patients in molecular subgroups is the main limitation of this work and may have led to a lack of power and sampling bias. Further limitations stem from its retrospective design, including reporting bias, a lack of centralized radiological assessment, and variable scanning intervals along with alpha inflation due to multiple testing.

Despite these limitations, this study suggests that the efficacy of combining pembrolizumab with pemetrexed and platinum-based chemotherapy differs according to certain histo-molecular biomarkers, which may help determine which patients are liable to benefit from CIT.

The histo-molecular response factors evaluated in our study, which are those currently available in clinical practice, reflect the biological diversity of NSCLC but do not seem sufficient for predicting the response to CIT. The identification of new biomarkers, especially immunological scores, could provide better understanding of the complexity of the tumor microenvironment and could help to optimize treatments through a personalized approach.