Abstract

Objectives

   Emerging data have shown that local treatment could provide clinical
   benefit for non-small cell lung cancer (NSCLC) patients with
   oligometastasis. Liver metastases have the worst prognosis in advanced
   NSCLC, but the genomic characteristics of liver oligometastasis remain
   unclear. The aim of our study was to elucidate the molecular features
   of liver oligometastatic NSCLC.

Methods

   Paired liver metastatic tissue samples and peripheral blood from 32
   liver oligometastatic NSCLC patients were concurrently collected for
   comprehensive genomic analysis using next-generation sequencing.

Results

   A total of 206 mutated genes in 32 patients were detected, with a
   median of 4 mutations per sample. The most frequent alterations (> 10%)
   in liver oligometastasis were TP53 (72%), EGFR (50%), RB1 (19%) and
   SMARCA4 (12%). The co-occurrence rate of TP53 and RB1 in our cohort was
   significantly higher than that in the TCGA-LUAD cohort. Age, APOBEC,
   homologous recombination deficiency (HRD) and deficient mismatch repair
   (dMMR) established the mutational signature of liver oligometastatic
   NSCLC. The median tumor mutation burden (TMB) was 4.8 mutations/Mb. A
   total of 78.12% patients harbored at least one potentially actionable
   molecular alteration that may guide further targeted therapy according
   to the OncoKB evidence.

Conclusions

   Our study comprehensively delineated the genomic characteristics of
   liver oligometastatic NSCLC - such findings were helpful to better
   understand the distinct clinic-biological features of oligometastasis
   and optimize personalized treatment of this population.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12885-025-13478-5.

   Keywords: NSCLC, Liver oligometastasis, Genomic profiling,
   Next-generation sequencing, Tumor mutational burden

Introduction

   Lung cancer has the highest cancer mortality and incidence in the world
   [[46]1, [47]2]. Approximately two-thirds of non-small cell lung cancer
   (NSCLC) patients present with metastatic disease at diagnosis. Liver is
   one of the most frequent sites of tumor metastases in advanced NSCLC
   patients, together with a poor overall survival (OS) [[48]3]. It is
   also considered an organ of immune tolerance, characterized by immuno-
   suppressive signals and T-cell anergy. With improved understanding of
   tumor biology, the oligometastatic state has been regarded as a unique
   subgroup. Liver oligometastasis refers to the metastases confined to
   the liver with a limited number of metastatic lesions as detected by
   computed tomography, magnetic resonance imaging and/or positron
   emission tomography-CT. Emerging evidence have demonstrated that local
   treatment, such as surgery or radiotherapy, could provide clinical
   benefit for NSCLC patients with oligometastasis [[49]4–[50]6]. However,
   little is known about the genomic characteristics of liver
   oligometastatic NSCLC. Since liver metastases remain the worst
   prognosis in advanced NSCLC, it is therefore important to investigate
   the molecular features of liver oligometastatic NSCLC, which will
   provide a better understanding basis for the biology of this distinct
   lung cancer subtype.

   The outcomes of NSCLC patients have been prolonged with the development
   of targeted therapies and immune checkpoint inhibitors (ICIs).
   Targetable mutant genes were often used to guide individualized
   targeted therapies under the tumor genome sequencing for primary lung
   cancer in the current age of precision medicine. However, the frequency
   and evidence level for actionable mutations in liver oligometastasis
   remains largely unclear since previous researches on targeted therapy
   have mostly focused on patients with conventional NSCLC. Moreover,
   tumor mutation burden (TMB) serves as a predictive biomarker for
   immunotherapy of multiple tumors [[51]7], but the TMB expression
   pattern in liver oligometastasis is still unknown. Thus, a
   re-evaluation on the actionability of targetable alterations and
   exploration of the predictive biomarkers of immunotherapy in liver
   oligometastasis could provide new insights into the clinical
   application of targeted therapy and ICIs among this patient population.

   In the present study, we report the unique genomic features of liver
   oligometastatic NSCLC and reveal its potential clinical significance in
   guiding therapeutic strategies to help clinicians in personalized
   cancer treatment selection.

Method and material

Study population and sample collection

   We enrolled 32 liver oligometastatic NSCLC patients in the Oncology
   Center of the Second Affiliated Hospital of Chongqing Medical
   University between February 2017 and September 2022. The major
   inclusion criteria for this study were: pathologically confirmed NSCLC,
   stage IV disease based on the 8th edition of the American Joint
   Committee on Cancer staging system, with only one metastasis confined
   to liver for three months or longer to ensure the true oligometastatic
   state. Involved regional lymph nodes were classified as the primary
   tumors and were not counted as metastatic lesions. Associated clinical
   and histopathological data were obtained from electronic medical
   records. The study was approved by the ethical committee of Chongqing
   Medical University. Written informed consent was obtained from each
   patient before inclusion.

DNA extraction and target capture sequencing

   Tissue specimens from 32 patients were sequenced by a 1021 gene panel
   in the Geneplus-Beijing Institute (Beijing, China) and were all taken
   from metastatic lesions. Gene lists of targeted sequencing panel may be
   found in supplementary Table [52]1. Genomic DNA was extracted from
   formalin-fixed and paraffin-embedded (FFPE) tumor tissues and matched
   peripheral blood using the QIAamp DNA FFPE tissue kit (Qiagen, Hilden,
   Germany). Then, the Qubit dsDNA BR assay (Life Technologies, USA) was
   used to quantify the extracted DNA. The custom-designed probes
   covering ~ 1.4 Mb of genomic sequences for 1021 cancer-related genes
   (Supplementary Table [53]1) were used to capture DNA. To construct the
   library, 1.0 µg of tissue DNA and matched peripheral blood were sheared
   into 300-bp fragments using a Covaris S2 ultrasonicator. Libraries were
   constructed using the KAPA DNA Library Preparation Kit. An Applied
   Biosystems 7500 real-time PCR system and an Agilent 2100 Bioanalyzer
   were used to measure the captured libraries. DNA sequencing was
   performed on the HiSeq3000 Sequencing System using 2 × 100 bp
   paired-end reads.

Sequencing data analysis

   From the raw sequencing data, adaptor sequences and low-quality reads
   were first removed. The human GRCh37 reference genome was used to
   perform the alignment using BWA (a Burrows-Wheeler aligner) [[54]8].
   PCR duplicates were marked by Picard tools
   ([55]http://broadinstitute.github.io/picard/). Variations were called
   by GATK (version 3.4-46-gbc02625) or MuTect (version 1.1.4) [[56]9,
   [57]10]. Germline mutations were filtered out by PBL sequencing. All
   candidate mutations were then reviewed manually with the Integrated
   Genome Viewer (IGV) [[58]11]. Variation sites were annotated with
   ANNOVAR software [[59]12].

Analysis of mutational signature

   Mutation signatures were characterized according to the six
   substitution patterns (T > A, T > C, and T > G, C > A, C > G, C > T)
   and 3′- and 5′- flanking nucleotides. The Catalogue of Somatic
   Mutations in Cancer (COSMIC) [[60]13] mutation signature was used as a
   reference to extract potential mutational signatures (R package
   MutationalPatterns) [[61]14]. Subsequently, the relative contribution
   of different mutational signatures in liver oligometastasis has been
   analyzed.

Clinical actionability based on OncoKB

   The Precision Oncology Knowledge Database (OncoKB) [[62]15] predicts
   drug actionability according to available clinical evidence, including
   regimens approved by Food and Drug Administration (FDA) and those still
   in clinical trials. This clinical support tool established an evidence
   classification system that could categorize potentially actionable
   mutations into different levels based on the evidence strength.
   Alterations with level 1–2 are FDA-approved or considered biomarkers
   that can predict the response to FDA-approved drugs. Level 3–4
   alterations signify compelling clinical or biological evidence that can
   predict the response to an existing drug. “Actionable mutations” are
   defined as genomic alterations corresponding to evidence level 1–4 in
   OncoKB that have therapeutic implications. Annotation of the genomic
   mutations using the OncoKB database was performed on January 29th,
   2023.

Statistical analysis

   The analyses of Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene
   Ontology (GO) enrichment were performed for all somatic mutations by
   the ClusterProfiler package [[63]16]. SPSS 26.0 software was used for
   statistical analysis. Categorical variables were compared by the
   Fisher’s exact or Chi-square test, while continuous data were compared
   by the Mann–Whitney U test or two-sample t test. Results were
   recognized as statistically significant when the two-sided p-value was
   less than 0.05.

Results

Patient characteristics

   Table [64]1 presented the clinical characteristics of the 32 liver
   oligometastatic NSCLC patients included in this study. Histologically,
   lung adenocarcinoma was diagnosed in all oligometastatic patients, and
   less than half of them were male (15/32, 46.9%). The median age of all
   included patients was 65 years (range, 34 to 88 years). The majority of
   patients were non-smokers (19/32, 59.4%), without a family history
   (27/32, 84.4%), with synchronous liver oligometastasis (24/32, 75.0%)
   and with a lymph nodal status of N2-N3 (23/32, 71.9%).

Table 1.

   Clinical characteristics of 32 liver-only oligometastatic NSCLC
   patients
   Characteristics               No. Of patients (%)
   Total                         32(100.0)
   Gender
    Male                         15(46.9)
    Female                       17(53.1)
   Age (years)
    Median (range)               65(34–88)
   Tumor histology
    Adenocarcinoma               32(100)
    Squamous cell carcinoma      0(0.0)
   Smoking
    Yes                          13(40.6)
    No                           19(59.4)
   Family history
    Yes                          5(15.6)
    No                           27(84.4)
   Type of liver oligometastasis
    Synchronous                  24(75.0)
    Metachronous                 8(25.0)
   Nodal status
    N0-1                         9(28.1)
    N2-3                         23(71.9)
   [65]Open in a new tab

Genomic characteristics of liver oligometastatic NSCLC

   The landscape of genomic mutations in 32 liver oligometastatic NSCLC is
   shown in Fig. [66]1A. The most frequent alterations (> 10%) in liver
   oligometastasis were TP53 (72%), EGFR (50%), RB1 (19%) and SMARCA4
   (12%). A total of 206 mutated genes in 32 patients were detected by
   targeted sequencing, with a median of 4 mutations per sample (range,
   0–24) (Fig. [67]1A). Among them, missense mutations were the most
   common type of mutation, followed by nonsense and frame shift deletions
   (Fig. [68]1B), while insertions and deletions occurred less frequently
   than single nucleotide polymorphisms (Supplementary Fig. [69]1A). We
   also described the CNV landscape of our cohort and found that the most
   frequent CNV changes were EGFR amplification (25.00%), followed by
   CDKN2A deletion (9.38%), CDKN2B deletion (9.38%), CDK4 amplification
   (6.25%) and MYC amplification (6.25%) (Supplementary Fig. [70]1B). For
   each sample, we counted and calculated the proportion of six-base
   substitution for the single nucleotide variation (SNV) spectrum
   (Supplementary Fig. [71]1D), and found that C > A transversion
   accounted for the most common substitution type in our patients
   (Fig. [72]1C), with a Ti/Tv ratio of 1.13 (Supplementary Fig. [73]1C).

Fig. 1.

   [74]Fig. 1
   [75]Open in a new tab

   (A) Mutation landscape of 32 liver oligometastatic NSCLC patients. The
   heatmap shows the top 20 genes across all samples, with genes ranked by
   mutation frequency. Top bar summarizes the total number of alterations
   in each patient (columns), and the dashed line indicates the median
   number of alterations. Side bar (rows) summarizes the percentage of
   tumors with mutation in each gene and mutation composition for each
   gene in the entire cohort. Bottom heatmap, smoking, gender and age
   information. Different colors denote different types of mutations and
   different clinical features. (B) Variant classification, (C) single
   nucleotide variations, (D) mutual exclusivity and co-occurrence
   analysis, (E) Comparison of the frequency of 20 high-frequency
   mutations identified in liver oligometastatic LUAD with that in the
   East Asian and TCGA LUAD. TCGA, The Cancer Genome Atlas. *p < 0.05, and
   **p < 0.01, ***p < 0.001

   Considering that all liver oligometastatic tumors in our cohort were
   lung adenocarcinoma (LUAD), we further compared the high-frequency
   mutated genes in our study with mutations in East Asian LUAD in
   Singapore Oncology Data Portal (OncoSG) cohort, Chinese LUAD cohort and
   TCGA LUAD cohort (Fig. [76]1E). The frequency of alterations in TP53
   and MLL3 in the liver oligometastatic group was found to be
   significantly higher than that reported in East Asian and TCGA LUAD.
   Notably, RB1 mutations were more frequent in patients with liver
   oligometastasis compared to all other cohorts (Fig. [77]1E). Because
   TP53 and EGFR were the most frequently mutations in this cohort and the
   subtypes of the two genes were strongly correlated with response to
   therapy and treatment option, they were analyzed separately to reveal
   the heterogeneity and mapped to corresponding protein sequences by
   MutationMapper. We found that the most frequently mutated sites of TP53
   and EGFR were Q331 and T790M (Supplementary Fig. [78]2A and B),
   respectively. And the mutation sites of the other top 20 highly mutated
   genes were also displayed in supplementary Fig. [79]2C to T. TP53 was
   the most common mutation in our cohort. A total of 25 mutations in TP53
   were identified among 72% (23/32) of the patients, and 44% of the
   alterations were truncated mutations leading to TP53 inactivation.
   Furthermore, we observed 16 patients (16/32 = 50%) harbored 41 EGFR
   mutations, of which EGFR p.L858R and exon 19del comprised 14.63% and
   17.07%, respectively (Supplementary Fig. [80]1E). EGFR/TP53
   co-alterations are correlated with shortened response to EGFR tyrosine
   kinase inhibitors (TKIs) and poor survival outcomes in EGFR-mutant
   NSCLC patients [[81]17–[82]19]. It is worth noting that the
   co-occurrence of EGFR and TP53 mutations accounted for 81.25% of
   EGFR-mutant oligometastatic tumors (Supplementary Fig. [83]1F).
   Previous studies have shown that the inactivation of both TP53 and RB1
   was associated with the histologic transformation of LUAD into
   small-cell lung cancer (SCLC). Of note, we found that that 5 (15.63%)
   of the 32 patients with oligometastatic LUAD had mutations in both TP53
   and RB1, which was significantly higher than that in the TCGA-LUAD
   cohort (15.63% vs. 4.81%, p = 0.026) (Supplementary Fig. [84]1G).
   Finally, we further performed the somatic interaction analysis to show
   the mutual exclusivity and co-occurrence of genetic alterations in
   liver oligometastasis. Mutations in KEL/GRIN2A, KEL/KEAP1, KEL/LRP1B,
   MLL3/PIK3CG, PTCH2/PIK3CG and PTCH2/MLL3 were significantly
   co-occurring (p<0.01), but mutually exclusive mutations were not found
   in our research (Fig. [85]1D).

KEGG and GO pathway enrichment analysis

   In order to disclose the significance and biological feature of mutated
   genes in liver oligometastasis, the enrichment analysis of KEGG and GO
   were conducted. Both somatic mutations and/or copy number alterations
   were analyzed for KEGG and GO pathway enrichment. The results were
   ranked based on the p value, and the top 20 enriched KEGG pathways
   (Fig. [86]2A) and GO (Fig. [87]2B) terms were presented in Fig. [88]2.
   Significantly altered pathways included central carbon metabolism in
   cancer, EGFR tyrosine kinase inhibitor resistance, PI3K-Akt signaling,
   ErbB signaling, FoxO signaling, Ras signaling, JAK-STAT signaling and
   other well-known pathways. Based upon these findings, we focused on the
   PI3K-AKT signaling because activation of this pathway may predict a
   higher risk of histologic transformation of LUAD to small cell lung
   cancer (SCLC) [[89]20]. A total of 90.6% (n = 29) of the patients
   carried alterations in genes of the PI3K-AKT pathway. In addition, we
   observed that the JAK-STAT signaling was significantly enriched in our
   cohort, which could alter the fibrotic and immune microenvironment of
   the liver to establish a pre-metastatic niche and strongly suppressed
   the antitumour immune response [[90]21]. 71.9% (n = 23) of liver
   oligometastatic patients harbored 54 genomic mutations in the JAK-STAT
   pathway. GO analysis revealed enrichment of genes were mainly related
   to kinase activity, phosphorylation and reproductive development.

Fig. 2.

   [91]Fig. 2
   [92]Open in a new tab

   Top 20 enriched pathways by (A) KEGG and functional terms by (B) GO
   enrichment of somatic mutations in liver oligometastasis. Count: the
   number of mutations enriched in this signaling pathway or functional
   term

Analysis of mutational signature and TMB

   In addition to genetic mutations and signaling pathways, the profile of
   mutational signature can help to better understand the specific
   mutagenesis process of liver oligometastatic tumors. To extract the
   potential mutational signatures from our tumor samples, we used the
   non-negative matrix factorization (NMF) approach to identify
   etiological mutational processes by decomposing the base substitution
   matrix. We found that the mutational signatures of oligometastasis
   included signature 1 A (40.0%), signature 2 (27.4%), signature 20
   (9.2%), signature 3 (16.6%) and unknown (4.8%) (Fig. [93]3A). Signature
   1 A characterized by C > T transitions contributed most to the
   mutational process in our cohort, which was proposed to be caused by an
   endogenous mutational process initiated by spontaneous deamination of
   5-methylcytosine and correlated with the age at cancer diagnosis. The
   particular feature of patients in this research whose median age was
   relatively old (median 65, range 34 to 88 years) may account for the
   abundance of signature 1 A. Signature 2 was related to APOBEC cytidine
   deaminase activity making up 27.4% of the observed signatures. Previous
   studies have demonstrated that the enrichment of APOBEC signatures in
   adenocarcinomas that may contribute to the transformation process of
   LUAD to SCLC [[94]22, [95]23]. Signature 20 was associated with
   defective DNA mismatch repair (dMMR), an abnormal DNA repair mechanism
   that ultimately lead to frequent genomic mutations and instability.
   Signature 3 contributing to 16.6% of the total signatures in our study
   was relevant to homologous recombination deficiency (HRD). The presence
   of multiple somatic mutations in our tumor samples, including BRCA2
   (n = 2, 6.3%) and RAD51C (n = 1, 3.1%), could support the DNA damage
   caused by HRD. The etiology of the remaining mutational signature
   (4.8%) in oligometastasis still remains unclear. In summary, age,
   APOBEC, HRD and dMMR established the mutational signature of liver
   oligometastatic NSCLC.

Fig. 3.

   [96]Fig. 3
   [97]Open in a new tab

   (A) Mutational signatures in liver oligometastatic NSCLC. (B) Density
   plot of tumor mutational burden (TMB) in all cancer patients. TMB
   according to (C) gender, (D) smoking status, (E) TP53 genotype, (F)
   EGFR genotype, (G) RB1 genotype and (H) SMARCA4 genotype

   TMB has been identified as a prognostic and predictive biomarker for
   immunotherapy in a variety of solid tumors. In our study, the median
   TMB of liver oligometastasis was 4.8 mutations/Mb (range 0–24.96
   mutations/Mb). The density plot in Fig. [98]3B shows the TMB
   distribution of all oligometastatic patients. Notably, we found that
   one patient had 24.96 mutations/Mb, significantly deviating from a
   normal distribution. Multiple alterations relevant to defective DNA
   repair genes, such as ATR, ERCC5, RECQL4 and TP53, were present in this
   sample. The median TMB in females was slightly higher than that in
   males (7.68 versus 3.84 mutations/Mb, p = 0.72) (Fig. [99]3 C).
   Meanwhile, for smoking patients, the median TMB was 7.68 mutations/Mb,
   which was higher than that in non-smoking patients (3.84 mutations/Mb,
   p = 0.29) (Fig. [100]3D). Since TP53 (72%), EGFR (50%), RB1 (19%) and
   SMARCA4 (12%) were the most common alterations (> 10%) in liver
   oligometastasis, we further analysed the underlying relationship
   between these high-frequency mutations and TMB level. Tumors with
   mutations in TP53, EGFR, or RB1 had higher median TMB than those with
   wild type (5.28 versus 1.86 mutations/Mb, p = 0.13; 4.8 versus 4.32
   mutations/Mb, p = 0.82; 5.76 versus 3.84 mutations/Mb, p = 0.51)
   (Fig. [101]3E, F and G). In addition, the median TMB of SMARCA4-mutant
   tumors was significantly higher than that of SMARCA4 wild-type tumors
   (17.28 versus 3.84 mutations/Mb, p = 0.04) (Fig. [102]3H).

Clinical actionability for targetable genes

   With the rapid development of targeted drugs and cancer genomics, a
   clinical re-evaluation on the significance of comprehensive genomic
   profiles for targeted therapy in patients with liver oligometastasis
   according to the actionability of specific genetic mutations
   categorized by the OncoKB database is warranted (Supplementary
   Fig. [103]3). A total of 78.12% patients harbored at least one
   potentially actionable molecular alteration that may guide further
   therapeutic strategy. Furthermore, 56.25% tumors carried mutations
   ranked as level 1 including missense mutations of ALK, BRAF, EGFR,
   ERBB2 amplification, ALK fusion and in-frame deletion of EGFR. The
   final evidence level for each sample was defined as the highest level
   of all alterations detected in the patient. Thus, patients were not
   assigned to level 2 and 3 because these patients would all have
   higher-level targets. Level 2 mutations include missense mutations of
   FGFR and ERBB2 amplification. Level 3 actionable alterations were SNVs
   of EGFR and amplification of ERBB2. Level 4 mutations accounted for
   21.88% including SNVs of BRAF, KRAS, PIK3CA and STK11, amplification of
   FGFR1 and deletion of CDKN2A (Fig. [104]4A, C, D). Drugs that target
   particular genomic mutations in our liver oligometastatic tumors were
   also presented in Fig. [105]4C. Given that better clinical outcomes can
   be achieved by combinations of various targeted drugs, we further
   profiled the distribution of patients carrying multiple actionable
   alterations. In general, 34.38% of liver oligometastatic patients
   harbored one actionable mutation, while 34.38% and 9.36% of those
   harbored two or more actionable mutations, respectively (Fig. [106]4B).

Fig. 4.

   [107]Fig. 4
   [108]Open in a new tab

   Somatic alterations identified by the 1021-panel that are clinically
   actionable. (A) Clinical evidence based on OncoKB was used to define
   alterations. (B) Patients were classified according to their highest
   level of actionable alterations (left). Mutations in different grades
   (right). (C) The percentage of patients with a single actionable
   mutation or multiple actionable mutations. (D) Distribution of levels
   of actionable mutations and their corresponding potential targetable
   drugs. (E) Distribution of alteration types of actionable genes

Discussion

   NSCLC has historically been considered as a heterogeneous entity with a
   high degree of genomic diversity, resulting in variable responses to
   therapy and clinical outcomes [[109]24]. Oligometastasis was recognized
   as a unique clinical disease with distinct molecular and
   clinic-biological features compared to widely metastatic disease, such
   as an indolent disease state that could benefit from local treatment.
   Liver metastases are common in advanced NSCLC patients, which could
   cultivate an immune desert that lead to inferior response to
   immunotherapy and shorter OS [[110]25]. Emerging evidence has
   demonstrated that NSCLC patients with liver oligometastasis could
   benefit from local treatment and achieve better progression-free
   survival or overall survival [[111]26, [112]27]. Until recently,
   current research has not reached a consensus on the exact number of
   metastases that define oligometastasis [[113]28]. Considering that most
   clinical studies enrolled oligometastatic patients with all metastatic
   lesions adding up to five or less [[114]29], we applied stricter
   inclusion criteria to include patients with only one distant metastasis
   confined to liver for three months or longer to ensure the true
   oligometastatic state. Comprehensive molecular profiling of cancers by
   next-generation sequencing methods is increasingly being used for
   therapeutic management decisions in oncology. Although rapid advances
   in local treatment have revolutionized cancer therapy in
   oligometastasis, little is known about the genomic characteristics of
   liver oligometastasis. Therefore, it is necessary to have a
   comprehensive and deeper understanding of molecular features of such
   patients to provide better insights into the underlying clinical and
   biological features of this disease.

   In our research, we found that the co-occurrence rate of TP53 and RB1
   in liver oligometastasis was significantly higher than that in the
   TCGA-LUAD cohort (15.63% vs. 4.81%, p = 0.026) (Supplementary
   Fig. [115]1G). The increase has profound clinical implications because
   previous studies have shown that the inactivation of both TP53 and RB1
   may facilitate the histologic transformation of LUAD into SCLC, with an
   approximately 43-fold increase in the risk of promoting transformation
   [[116]22]. In particular, the prognosis of transformed SCLC is
   significantly poor, even worse than de novo SCLC [[117]30]. We also
   noticed that PIK3CA (9%) was one of the high-frequency mutations and
   90.6% (n = 29) of the patients carried genomic alterations in the
   PI3K-AKT pathway. Emerging data also support that upregulation of
   PIK3CA gene mutation and PI3K/AKT signaling occurred earlier during the
   neuroendocrine transformation [[118]20]. In addition, mutational
   signature 2 related to APOBEC (27.4%) as one of the dominant signatures
   in our cohort may promote the transformation process and drive drug
   resistance to EGFR TKIs [[119]31]. These suggest that small cell
   histologic transformation occurs frequently in liver oligometastasis
   with a worse prognosis. Therefore, patients at the initial diagnosis of
   liver oligometastasis should be advised to undergo comprehensive
   genetic testing to clarify the presence of concurrent RB1 and TP53
   alterations in order to provide guidance for future tumor evolution.
   Tumor tissues should also be retaken for pathology in time at the early
   stage of disease progression to determine whether the histological
   transformation occurs. Alpelisib, a PI3K-AKT inhibitor that
   significantly inhibits both TP53/RB1 expression in patient-derived cell
   model [[120]32], can delay tumor growth in NSCLC undergoing histologic
   transformation. It may be a potential treatment option for liver
   oligometastatic NSCLC in future. However, these effects are currently
   the results of preclinical studies, and additional clinical trials are
   needed to validate the efficacy and safety of Alpelisib in inhibiting
   SCLC transformation in the real world.

   TP53 was the most frequently mutated gene in our study and might have
   negative prognostic effects for NSCLC [[121]33]. It co-occurred with
   EGFR mutations to reduce responsiveness to EGFR TKIs and was associated
   with poor survival outcomes [[122]17–[123]19]. The incidence of
   concurrent TP53/EGFR in liver oligometastasis (81.25%) was higher than
   that of EGFR-mutated NSCLC (55–65%) in other researches. Given the high
   ratio of concurrent TP53/EGFR, a role for APOBEC mutagenesis in the
   development of resistance to targeted therapies and the enrichment of
   EGFR tyrosine kinase inhibitor resistance signaling pathway, EGFR TKIs
   might have low efficiency in this particular population. However,
   preclinical studies indicate that inhibiting APOBEC mutagenesis through
   gene deletion or RNA interference could delay resistance to targeted
   therapies in lung cancer cell lines [[124]34]. Although direct
   inhibitors of APOBEC activity are currently unavailable, prior research
   has suggested a potential role for ataxia telangiectasia and
   Rad3-related (ATR) and PARP inhibitors [[125]35, [126]36].
   Specifically, studies have shown that activation of APOBEC may
   sensitize tumor cells to inhibitors of the DNA damage response pathway,
   such as ATR and PARP inhibitors. In vitro studies have demonstrated
   that ATR inhibition can effectively overcome EGFR-TKIs resistance
   [[127]37]. Therefore, these findings suggest that DNA damage repair
   inhibitors in combination with EGFR-TKIs may hold promise for
   overcoming resistance and enhancing therapeutic efficacy in liver
   oligometastatic patients.

   Patients with liver metastases have a lower response rate to
   immunotherapy and shorter survival time than those with metastases to
   other organs in lung cancer [[128]25]. Of all the pathways enriched in
   oligometastasis, we focused on the JAK-STAT signaling signaling because
   activation of this pathway could alter the fibrotic and immune
   microenvironment of the liver to establish a pre-metastatic niche and
   lead to a highly immunosuppressive tumour microenvironment that
   severely hindered antitumour immunity [[129]21]. Recent studies have
   shown that blockade or genetic ablation of components of JAK-STAT
   signaling could prevent establishment of a pre-metastatic niche and
   inhibit liver metastasis [[130]21]. However, the PI3K pathway promotes
   tumour progression and growth independently of JAK-STAT3. Inhibitors of
   JAK-STAT3 signaling may not be effective as monotherapy in tumours with
   already activated PI3K pathway. Therefore, JAK-STAT3 inhibitors
   combined with PI3K signaling pathway inhibitors will be an effective
   therapeutic regimen for liver oligometastasis.

   TMB is a potential biomarker used to predict treatment response to ICIs
   in NSCLC. Tumours with liver oligometastasis in our study had a lower
   median TMB (4.8 mutations/Mb) compared with previously reported brain
   oligometastatic NSCLC (8.7 mutations/Mb). This finding is consistent
   with previous study that TMB is a site-specific biomarker correlated
   with tissue location and brain metastases have the highest TMB values
   compared to metastases in other organs in NSCLC [[131]38]. Furthermore,
   SMARCA4 mutation was significantly positively associated with TMB in
   our cohort, similar to other study in conventional NSCLC [[132]39].
   Previous studies have shown that SMARCA4-mutant NSCLC tend to derive
   significant benefit from ICI treatment [[133]39]. Although liver
   metastasis systemically restrains immunotherapy efficacy and remains an
   independent predictor of poor response to PD-(L)1 blockade, our study
   suggests that SMARCA4-mutant patients, an important subgroup of liver
   oligometastasis, may be more sensitive to and benefit well from
   immunotherapy compared with the overall population. Several researches
   have suggested SMARCA4 mutations were potential targets for lung
   cancer, but there are currently no effective targeted therapies for
   SMARCA4-mutant NSCLC. However, CDK4/6 inhibitors have recently been
   reported to show antitumor activity in SMARCA4 deficient tumors, so
   palbociclib (CDK4/6 inhibition) could be a promising option for the
   treatment of oligometastasis [[134]40]. Notably, preclinical data
   suggest that increased ERK-driven mTOR pathway signalling is associated
   with resistance to palbociclib in lung cancer cell lines and upstream
   inhibition with a ERK inhibitor plus palbociclib increases cell
   apoptosis [[135]41]. Based on these promising preclinical data,
   clinical trials are currently ongoing in advanced NSCLC investigating
   CDK4/6 inhibitors in combination with ERK inhibitors ([136]NCT02857270,
   [137]NCT03454035). Concurrent CDK4/6 and ERK inhibition may be more
   efficacious than CDK4/6 inhibitor monotherapy, which may be a better
   treatment option for liver oligometastatic patients.

   With the discovery of novel targetable driver genes and the emergence
   of matched new targeted drugs, the clinical prognosis of lung cancer
   patients with specific gene mutations has been greatly improved. In the
   present study, we detected a total of 78.12% liver oligometastatic
   patients carried at least one potentially actionable alteration that
   may guide further treatment strategy based on OncoKB evidence. This
   proportion was higher than the 67% reported in a previous study that
   included 1564 conventional advanced NSCLC patients [[138]42],
   indicating genomic mutations in liver oligometastasis are highly
   targetable. Targeted therapy matching level 1–2 genetic alterations
   could significantly extend progression-free survival and OS in NSCLC,
   but marked clinical advantage was not observed in patients with level
   3–4 alterations [[139]42]. The percentage of patients with level 1–2
   and 3–4 mutations as their highest actionable targets was comparable to
   usual advanced NSCLC patients, respectively (56.25% vs. 57.1%; 21.88%
   vs. 19.2%).

   In conclusion, our study comprehensively delineated the genomic
   characteristics of liver oligometastatic NSCLC – such findings were
   helpful to better understand the distinct clinic-biological features of
   these patients. We have also discussed promising and suitable
   therapeutic strategies for liver oligometastasis to optimize
   individualized treatment of this population. Finally, there are some
   limitations that require mentioning. First, we diagnosed
   oligometastasis in the absence of a precise diagnostic standard, so we
   might include polymetastatic patients. Second, the inclusion criteria
   for liver oligometastatic samples were more stringent than in other
   clinical studies, resulting in a limited sample size (n = 32). Third,
   this research was a retrospective study with risks of selection bias.
   Fourth, TMB was calculated using panel sequencing, which might be less
   accurate compared with whole-exome sequencing. Fifth, while 1021-panel
   sequencing covers a large number of genes, targeted panels can still
   miss certain alterations due to technical limitations. In contrast,
   whole exome sequencing or whole genome sequencing would allow for more
   comprehensive analysis to further improve the accuracy and
   applicability of the results.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [140]12885_2025_13478_MOESM1_ESM.docx^ (2.4MB, docx)

   Supplementary Material 1: Supplementary Fig. 1. (A) variant type, (B)
   prevalence of CNV alterations, (C) Ti/Tv ratios, (D) Mutational
   fraction of the six-base substitution for each sample, (E) Frequency
   distributions of EGFR, (F) Frequency of concurrent EGFR/TP53 mutations,
   (G) Frequency of concurrent RB1/TP53 mutations in liver
   oligometastasis. Supplementary Fig. 2. The mutated sites of genes with
   high mutation frequency, including TP53 (A), EGFR (B), RB1 (C), SMARCA4
   (D), BRAF (E), GRIN2A (F), KEAP1 (G), LRP1B (H), PIK3CA (I), PIK3CG
   (J), RBM10 (K), BRCA2 (L), CDKN2A (M), KEL (N), KRAS (O), MLL3 (P),
   MSH6 (Q), PTCH2 (R), SETD2 (S), SMAD4 (T). Supplementary Fig. 3.
   Clinical evidence based on OncoKB was used to define alterations
   [141]12885_2025_13478_MOESM2_ESM.xlsx^ (15.6KB, xlsx)

   Supplementary Material 2: Supplementary Table 1. Gene list of the
   1021-gene panel

Acknowledgements