Somatic genomic alterations in individual genes may modulate the efficacy of PD-1 or PD-L1 inhibitors across various tumour types (Schoenfeld et al. 2020; Wu et al. 2022). KRAS gene mutations are one of the most common actionable genetic mutations in NSCLC, affecting up to 30% of patients (Hammerman et al. 2012; Collisson et al. 2014; Cascetta et al. 2022). Approximately 40% and 21% of all KRAS mutant cases carry two of the most common substitutions in codon 12 (G12C and G12V) (Reck et al. 2021). According to the literature, approximately 8.6% of NSCLC patients without KRAS mutation and 25% of KRAS-mutated NSCLC patients are carriers of STK11 mutations (Judd et al. 2021; Krzyżanowska et al. 2023). In KRAS-mutated NSCLC patients, the additional presence of an alteration in STK11 and KEAP1 genes decreased the response rate to immunotherapy, while their impact on ICI efficiency in the KRAS wt cohort is unknown (Ricciuti et al. 2022). Moreover, the presence of a STK11 mutation is associated with a poor prognosis in NSCLC (Rosellini et al. 2022).

In our cohort, KRAS mutations in codon 12 (G12C and G12V) were observed in 20% of patients (4/20). The STK11 gene was mutated in 15% of patients from our study group (3/20), which is consistent with other literature data (Facchinetti et al. 2017; Fleur et al. 2019; Bange et al. 2019). Furthermore, in our cohort, all 3 STK11 mutations overlapped with the KRAS mutation. In one case, it was associated with rapid progression during immunotherapy, while one patient surprisingly presented a very long response (18 months and still going on). Long responders with KRAS and STK11 mutations also carried alterations in the RAD50 and PMS2 genes. In contrast, the third patient with coexisting mutations in the KRAS and STK11 genes achieved a four-month stabilisation and had abnormalities in the ATM and PMS2 genes. RAD50 protein is involved in DNA double-strand break repair, and PMS2 protein in mismatch DNA repair, while ATM protein plays a significant role in homologous DNA repair (Abraham 2001; Bian et al. 2019). The loss of function of these proteins results in the accumulation of DNA damage and an increase in several somatic mutations in tumour cells (high TMB). A high number of neoantigens is the cause of the high immunogenicity of tumours (“hot” tumours) and may result in greater effectiveness of immunotherapy. Tumours harbouring concurrent KRAS/STK11 and KRAS/KEAP1 mutations display distinct immune profiles in terms of gene expression and immune cell infiltration (Rosellini et al. 2022), which may be associated with differences in response to immunotherapy. The immunological differences between 3 patients with concurrent KRAS/STK11 mutations are described in the section ‘Special alterations – KRAS and STK11 co-mutation.

According to TCGA data, the TP53 gene is the most frequently mutated in the LUAD cohort, and 30% of our patients (6/20) also had an alteration in this gene. However, in our study, alterations of KDR and NOTCH4 genes were the most common and affected 40% of patients (8/20), which was higher than reported in the TCGA data. In a pan-cancer study, KDR mutation was associated with higher TMB and immune cellular activities, indicating an immune-hot status characterised by higher PD-L1 expression and abundance of cytotoxic lymphocytes, which might suggest that KDR mutations are a potential positive biomarker for the long-term response to ICIs (Cui et al. 2022). In our study, KDR alterations were observed in both long and short responders, and this was not related to PD-L1 expression. The second most frequently altered gene in our cohort was NOTCH4, which was more common in short responders than in long responders. However, inflammatory TME in patients with a high-mutated NOTCH4 signalling pathway may be associated with a better prognosis in those receiving ICIs therapy (Lin et al. 2019; Li et al. 2021b). NOTCH4 mutations were associated with better response in NSCLC patients, and the NOTCH4 alterations overlapped with high TMB (> 10 mutations/megabase) (Long et al. 2021). Indels are the most common type of NOTCH4 alterations (Zhang et al. 2020). Moreover, co-occurring mutations in the NOTCH4 gene and genes encoding proteins in homologous repair pathways are associated with an increased efficacy of immunotherapy in advanced NSCLC patients (Mazzotta et al. 2020). In our cohort, most of the NOTCH4 alterations were indels. However, we noted a low TMB level that may be related to the specific targeted sequencing approach we used. Our cohort was too small to calculate the correlation between the co-occurrence of mutations and alterations in cell signalling pathways.

Furthermore, in our cohort, 4 mutations in the H3-3 A gene were observed only in the short-responders group. These alterations were present in both NOTCH4 and KDR-mutated patients, which may decrease the predictive impact of alterations in the NOTCH4 and KDR genes. The alterations of the H3-3 A gene affect the histone H3.3, whose high expression is associated with lung cancer progression by promoting cell migration and the activation of metastasis-related genes (Park et al. 2016; Bryant et al. 2020). The short responders’ group also had the WNT signalling pathway enriched, which drives the carcinogenesis and development of distant metastases (Zhan et al. 2016). Moreover, WNT signalling affects tumour immune surveillance, increasing resistance to immunotherapy (Zhou et al. 2022a).

Concerning immunological features, long responders had a significantly higher percentage of baseline CD4+/GATA3+ cells and a higher baseline Treg percentage than short responders. GATA3 promotes the differentiation of T helper (Th) cells into Th2 cells (Zhu et al. 2006). Th2 lymphocytes act antagonistically to Th1 cells, and the latter are crucial in anti-tumour immunity, i.e. by stimulating the activation of T cytotoxic (Tc) cells (Basu et al. 2021). Then, Treg also inhibits anti-cancer immunity and contributes to tumour growth. Those two parameters indicate the depletion of the immunological response. This finding is somewhat inconsistent with existing literature data, which suggest that a lower baseline Treg proportion predicts a better response (Bai et al. 2020; Li et al. 2021a). However, supposedly greater exhaustion in certain subpopulations results in greater activation after PD-1 blockade, as high expression of PD-L1 predicts better response. In Kang et al.’s study, the proportion of Tregs decreases after treatment in responsive patients, and this could also happen in our cohort (Kang et al. 2022).

In our cohort, patients with the KDR pathogenic variant had a higher proportion of PD-1+ Tc cells than the KDR-wt group. Alterations of KDR, as highlighted earlier, are related to higher PD-L1 expression and Tc abundance, with KDR activation being relevant in angiogenesis and metastases. Vascular endothelial growth factor A (VEGFA), a ligand to KDR, may enhance inhibitory checkpoint expression on CD8+ T cells (Voron et al. 2015). Moreover, we observed lower CXCR3 expression (one of the chemokine receptors) on CD4+ T cells and their lower ability to produce IFN-γ. CXCR3 is a receptor interacting with CXCL10 and CXCL9 chemokines. Notably, these chemokines inhibit angiogenesis (Cannon et al. 2021). Additionally, CXCR3 supports Th1 cells (Groom et al. 2012). However, CXCR3’s role is ambiguous and depending on many factors, it may be either inhibition or promotion of tumour growth (Chow et al. 2019; Russo et al. 2020). A lower percentage of IFN-γ-producing CD4+ cells is unfavourable in terms of anti-cancer response and speaks for a quiescent immune system (Jorgovanovic et al. 2020). Considering the above, the KDR role is still to be determined, especially since pathogenic variants have been found in both long and short-responders. Interestingly, the Notch pathway is also engaged in Th2 development through the regulation of GATA3 expression (Fang et al. 2007). In the absence of GATA3, Notch may promote Th1 differentiation (Amsen et al. 2007). Therefore, genetic alterations influence immunological processes in many ways.

In the Kaplan-Meier analysis, in patients with irAEs, we observed shorter PFS than in patients without adverse events, and this is because it was a reason for immunotherapy termination in case of severe events (PFS time was calculated to the end of immunotherapy). More extended observation of the patients is undoubtedly required. The lack of NOTCH4 alterations turned out to be favourable; however, as mentioned above, if pathogenic variants occurred, they often overlapped with other, not entirely beneficial alterations. Surprisingly, a high percentage of CD4+/GATA3+ cells was favourable in terms of PFS. As GATA3-derived Th2 cells are described as detrimental in cancer, it is not understood why a higher Th2% is beneficial. High CD62L expression on CD8+ cells’ positive influence may be explained by enhanced diapedesis of Tc lymphocytes to and from the lymph nodes, where the tumour antigens are presented.

As mentioned above, certain co-mutations that overlap with KRAS alterations modify the outcome of ICI therapy. Increasing evidence suggests that KRAS/STK11-mutated patients respond poorly to ICIs. Nevertheless, two out of three KRAS/STK11 mutated individuals in our study had at least 4 months of stabilisation. Those two patients share alterations in genes coding proteins involved in DNA repair (PMS2 in both patients, RAD50 in one patient, and ATM in the other). PMS2 encodes a key component of the mismatch repair (MMR) system, which aims to correct DNA mismatches and small insertions or deletions (Jenkins 2009). The altered MMR pathway is known to predispose colorectal cancer (and other solid tumour) patients to respond to ICIs, and it is connected to high microsatellite instability (MSI) and neoantigen production (Le et al. 2017; Overman et al. 2018). Homologous recombination deficiency (HRD) has also been described as a cause of genomic instability or neoantigenesis and as a possible predictor of clinical benefit during immunotherapy in pan-cancer and lung cancer studies (Yang et al. 2022; Khaddour et al. 2022; Zhou et al. 2022b). Interestingly, STK11 loss-of-function mutations cause suppression of cyclic GMP-AMP synthase (cGAS) Stimulator of Interferon Genes (STING) pathway. This signalling pathway is responsible for inducing inflammation via IFN expression after sensing cytosolic double-strand DNA (dsDNA), a physiologically relevant process during microbial infection (Kitajima et al. 2019). The cGAS-STING pathway may also be activated by genomic instability (Li and Chen 2018). STK11 mutations may alter the cell reaction to DNA damage and cause immunosuppressive changes in the tumour microenvironment (TME). On the other hand, the STING pathway may be activated in DNA repair deficiencies, suggesting that STK11 mutations and DNA repair alterations act in an antagonistic manner (Parkes et al. 2017). Some authors also highlight non-STING-dependent pathways for IFN signalling and inflammation after DNA damage (Zhang et al. 2019). Patients with both KRAS/STK11 alterations and DNA repair gene mutations achieved more durable clinical benefits; however, one alteration may abolish the effect of the other. Additionally, the flow cytometry results from the patient with rapid progression deviated from those of the other two patients with KRAS/STK11 mutations. The increase in the percentage of T lymphocytes expressing TIM-3 and LAG-3 was noticeable, and it has been reported that the expression of negative immune checkpoints, other than PD-1, can contribute to resistance to immunotherapy (Thommen et al. 2015; Wang et al. 2020). It appears possible that distinct mutations influence the immune system and shape the response to immunotherapy.

If FACS parameters are included in the analysis, the following molecules appear to be relevant: CD62L, GATA3, CTLA-4, TIM-3, LAG-3, CXCR3, PD-1, and EOMES. Nevertheless, their role is largely unevaluated, as there is limited research on these molecules regarding their response to ICIs. Some genetic alterations and certain molecules detected in cytometric analysis demonstrate predictive potential (Li et al. 2021a; Kwiecień et al. 2022; Miao et al. 2022; Ichiki et al. 2022).

This study has several limitations. First, the relatively small sample size introduces a potential bias; therefore, further research is currently being conducted to validate these findings in a larger cohort, which is expected to provide stronger statistical power. Moreover, we will incorporate dynamic changes in genetic and immunological parameters into future research, as this study only examined the parameters prior to treatment. Most studies on immune cell phenotype in NSCLC have focused on TILs; our analysis was based on PBMCs, which may account for certain differences and complicate direct comparison. Finally, the study group may be considered heterogeneous in terms of treatment type, age, PD-L1 expression, and other clinical characteristics, which may have influenced the observed variability in immune profiles.

As described in the discussion section, genetic abnormalities influence the immune landscape in cancer; therefore, it appears that assessing concurrent immunological and genetic changes is necessary to understand the anti-cancer response and discover new predictive factors. A growing body of scientific evidence suggests that a single biomarker may not be entirely sufficient in predicting such a complex process as a response to immunotherapy. Instead, it would be reasonable to create a whole diagnostic panel covering genes of proven significance. In our study, some of the alterations detected appear to be favourable, whereas others seem disadvantageous, and this is not always consistent with the literature data. Nevertheless, there is much more research to be done in this matter to determine the most essential alterations and their consequences. The best approach to predict response or resistance to immunotherapy remains elusive. Due to the limited number of samples in a studied cohort, we often refer our results to larger cohorts such as TCGA. Although our results are mostly consistent with previous reports, they should be confirmed in the larger study.