Spatial Immunogenomic Patterns Associated with Lymph Node Metastasis in Lung Adenocarcinoma

doi:10.21203/rs.3.rs-4827424/v1

Download PDF

Research Article

Spatial Immunogenomic Patterns Associated with Lymph Node Metastasis in Lung Adenocarcinoma

https://doi.org/10.21203/rs.3.rs-4827424/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background

Lung adenocarcinoma (LUAD) with lymph node (LN) metastasis is associated with poor prognosis, yet the specific mechanisms involved remain unclear. The objective of this investigation is to elucidate the immunogenomic landscape associated with LUAD with LN metastasis.

Methods

We utilized broad-panel next-generation sequencing (NGS) on a cohort of 257 LUAD patients who underwent surgical treatment. This approach allowed us to understand the molecular landscape of tumors and identify targetable driver-gene alterations. We also employed multiplex immunohistochemistry (mIHC) on the propensity score matching cohort, which enables comprehensive profiling of the tumor immune microenvironment while preserving cellular metaclusters, interactions and neighborhood functional units. By integrating data from both NGS and mIHC, we not only discerned spatial immunogenomic patterns within this meticulously matched cohort but also developed and independently validated a predictive model for LN stage.

Results

Our analysis revealed distinct patterns of immunogenomic alterations correlated with LN metastasis stages. Specifically, increased mutation frequencies in genes such as PIK3CG, ATM, BRD4, and KMT2B were observed alongside LN metastasis. Additionally, an enrichment of macrophages and regulatory T cells was associated with the immunogenomic patterns. Furthermore, a novel predictive model for LN metastasis likelihood was developed, offering potential benefits for patients ineligible for surgery.

Conclusions

This study offers an in-depth analysis of the genetic and immune profiles in LUAD with LN metastasis, identifying key immunogenomic patterns linked to metastasis. The creation of a predictive model from these insights marks a critical advancement in personalized treatment, underscoring its promise for enhancing patient management.

lung adenocarcinoma lymph node metastasis

genomic signatures

tumor immune microenvironment

spatial immunogenomic patterns

Lung adenocarcinoma (LUAD) metastasis, especially lymph node (LN) metastasis, plays a crucial role in influencing the clinical management and outcomes of patients [1–3]. However, our understanding of the intricate molecular mechanisms and evolutionary dynamics underlying lung cancer metastasis is limited due to the substantial genomic and transcriptomic diversity observed within and between patients [4]. Additionally, the scarcity of large-scale proteogenomics studies in Eastern populations, specifically focusing on primary tumor samples, hampers our ability to comprehensively characterize metastatic processes. On the other hand, studies that relied on biopsy samples face limitations in studying the driving events of metastasis due to the spatially altered clonal or subclonal structure and molecular landscape. Furthermore, unlike colorectal cancer [5] and breast cancer [6], which have seen extensive research on elucidating metastatic trajectories, the corresponding patterns in lung cancer metastasis remain largely unexplored.

Broad-panel next-generation sequencing (NGS) has revolutionized the field of oncology by providing insights into the molecular landscape of tumors, facilitating the identification of targetable driver-gene alterations, and informing prognoses. However, the specific genomic factors associated with LN metastasis in East Asian patients with LUAD have not been thoroughly investigated. Previous studies conducted in Western cohorts have emphasized the importance of genes, including STK11/SMACA, in LN metastasis [7]. To address this knowledge gap and enhance our understanding of LN metastasis in Chinese patients with LUAD, we examined the genomic characteristics and their correlation with pathologic LN status in patients who underwent surgical treatment.

The tumor immune microenvironment (TIME) contributes significantly to the heterogeneity of LUAD and impacts tumor progression as well as LN metastasis. The spatial organization of immune cells within the TIME affects their functionality and the impact on the tumor ecosystem. Although single-cell analysis techniques like scRNA-seq and CyTOF have been employed to examine the cellular composition of tumors, they lack the required spatial resolution to comprehend the spatial distribution and functional relationships among different metaclusters within the TIME[8]. Multiplex immunofluorescence (mIHC) provides a practical solution by enabling the simultaneous detection of multiple immune markers while preserving the spatial context, thus offering a spatially resolved single-cell phenotype [9, 10]. However, a comprehensive dimensional characterization of LN metastasis in LUAD through mIHC has yet to be explored.

The progression of LN metastasis in LUAD encompasses intricate spatial processes, encompassing tissue reorganization, invasion, and metastasis [11]. Previous investigation has highlighted the interplay between the genome and transcriptome in influencing LUAD evolution and metastasis [12]. Nevertheless, the exploration of metastasis progression requires a comprehensive understanding of spatial multi-omic patterns that are yet to be fully revealed. The TIME comprises a dynamic ecosystem consisting of tumor cells and their surrounding cellular components. Integrative spatial genomic and proteomic technologies present novel opportunities to explore cancer evolution at both molecular and spatial dimensions. Utilizing NGS and mIHC, we identified spatial immunogenomic patterns that offer detailed characterizations of the TIME, cellular interactions, and micro-anatomical structures. Spatial proteogenomics exposed the distinctive relationship between tumors positive for LN involvement and their microenvironments, providing insights into the underlying mechanisms that drive heterogeneity and opening new avenues for investigating LN metastasis in LUAD.

Therefore, we utilized a combination of NGS and mIHC to conduct an integrative analysis of an Eastern cohort consisting of patients with LUAD in this study. Through the identification of spatial immunogenomic patterns, we performed a functional analysis of topological cellular neighborhoods (CN), which unveiled unique spatial interactions and associations with specific gene alterations to reveal the immunogenomic pattern of the dynamic LUAD ecosystem. Based on this, we developed a novel model with integrated immunogenomic characteristics to predict the clinical LN stage for patients with LUAD who are not eligible for surgery. These findings contribute to an improved understanding of the pathophysiological mechanisms involved in LN metastasis and emphasize the potential for personalized treatment strategies that target immunogenomic patterns in LUAD.

Genomic Features Associated with Pathologic LN Metastasis in Eastern Asian LUAD

To identify genomic features associated with pathologic LN metastasis in LUAD primary tumors, we conducted the broad panel NGS analysis on the Eastern cohort. The study included a total of 257 patients who met the inclusion criteria (Table 1), with 54.8% (n = 141) were female and 71.6% (n = 184) were never smokers. Diagnostic CT scans revealed a solid morphologic appearance in 53% (108/204) of pN negative tumors and 90% (48/53) of pN positive tumors, showing a significant difference (p value < 0.001). The total tumor size, as measured on CT scans, was significantly larger in pN positive tumors (median interquartile range [IQR], 29 [23–36] mm) compared to pN negative tumors (median [IQR], 23 [16–31] mm; p value < 0.001). The majority of patients (n = 193 [75%]) had clinical stage I disease, while 22% (n = 10) had clinical stage II disease. Pathologic review revealed that 79% (n = 204) of resected tumors were pN negative, while 21% (n = 53) were pN positive. The ACI/PAP predominant histologic subtype was significantly lower in pN positive tumors compared to pN negative tumors (59% vs. 23%, p value < 0.001). The ACI/PAP predominant histologic subtype was significantly lower in pN positive tumors (59%) compared to pN negative tumors (23%, p value < 0.001). Lymphovascular invasion and visceral pleural invasion were significantly higher in pN positive tumors (38/53 vs. 52/204, 30/53 vs. 35/204 for pN positive vs. pN negative, respectively, P < 0.001). The clinicopathologic features of the patients are summarized in Table 1.

Table 1

Clinicopathologic characteristics of LN metastasis
Characteristics	pN negative	pN positive	p value
n	204	53
Age, median (IQR)	61 (54, 68)	63 (56, 68)	0.548
Sex, n (%)			0.775
Male	93 (45.6%)	23 (43.4%)
Female	111 (54.4%)	30 (56.6%)
Smoking, n (%)			0.985
No	146 (71.6%)	38 (71.7%)
Yes	58 (28.4%)	15 (28.3%)
Nodule_type, n (%)			< 0.001
solid	108 (52.9%)	48 (90.6%)
subsolid	96 (47.1%)	5 (9.4%)
Histologic subtype, n (%)			< 0.001
ACI/PAP	152 (74.5%)	23 (43.4%)
MIP/SOL	30 (14.7%)	29 (54.7%)
LEP	7 (3.4%)	0 (0%)
Others	15 (7.4%)	1 (1.9%)
Path_STAS, n (%)			0.262
no	181 (88.7%)	44 (83%)
yes	23 (11.3%)	9 (17%)
Path_VPI, n (%)			< 0.001
no	169 (82.8%)	23 (43.4%)
yes	35 (17.2%)	30 (56.6%)
Path_LVI, n (%)			< 0.001
no	152 (74.5%)	15 (28.3%)
yes	52 (25.5%)	38 (71.7%)
Stage, n (%)			< 0.001
IB	49 (24%)	0 (0%)
IIIA	2 (1%)	32 (60.4%)
IA2	79 (38.7%)	0 (0%)
IIB	4 (2%)	17 (32.1%)
IA1	33 (16.2%)	0 (0%)
IA3	32 (15.7%)	0 (0%)
IIA	5 (2.5%)	0 (0%)
IIIB	0 (0%)	4 (7.5%)
Note: Data are number (%) or median (interquartile range). ACI, acinar; LEP, lepidic; LVI, lymphovascular invasion; PAP, papillary; MIP, micropapillary; SOL, solid; STAS, spread through air spaces; VPI, visceral pleural invasion.

In the entire cohort, EGFR was the most commonly mutated gene in pN negative tumors (69%), whereas it was altered in 57% of pN positive tumors (Fig. 1). TP53 alterations were significantly more frequent in pN positive tumors than in pN negative tumors (55% vs. 37%; p value = 0.02, Fig. 1A). Notably, PIK3CG alterations were significantly more frequent in pN negative tumors compared to pN positive tumors in the NGS cohort (7.5% vs. 0.5%; p value < 0.05).

We performed subsequent analysis to examine genomic differences in the entire cohort of LN metastasis, focusing on the tumor mutation count, tumor mutational burden (TMB), and mutant-allele tumor heterogeneity (MATH) score (Fig. 1B-D). The results demonstrated a significant disparity in the tumor mutation count between pN positive and pN negative tumors (median [IQR], 6 [3–9] vs. 4 [2–6]; p value = 0.037; Fig. 1B). The TMB was notably higher in pN positive tumors (median [IQR], 4.74 [2.63–7.38] mutations/Mb) compared to pN negative tumors (median [IQR], 3.16 [1.75–4.92] mutations/Mb; p value = 0.017; Fig. 1C). The MATH score exhibited a significantly higher value in pN positive tumors (median [IQR], 53.49 [19.03–72.51]) compared to pN negative tumors (median [IQR], 45.67 [8.81–54.45]; p value = 0.02; Fig. 1D), suggesting a relatively elevated level of intratumor heterogeneity in pN positive tumors.

In the entire cohort of broad-panel NGS, we identified 13 driver genes with a false discovery rate (FDR) of 0.1. The most commonly detected drivers were EGFR (66%), TP53 (41%), and KRAS (12%). Notably, pN positive tumors exhibited a significantly higher frequency of TP53 mutations than pN negative tumors (54% vs. 37%, q value for Fisher’s exact test: < 0.001; Fig. 1E). Moreover, there were no statistically significant differences between pN negative and pN positive tumors in the prevalence of EGFR, KRAS, RBM10, RB1, STK11, SETD2, SF3B1, PTEN, HGF, BRAF, ARID1A and KEAP1 mutations, which were mainly driven by the entire cohort.

The present study investigated the co-occurrence and mutual exclusivity patterns (COME) of EGFR and several key genes in a large panel of NGS data. Our analysis revealed that EGFR-RBM10, KRAS-STK11, KEAP1-STK11, KEAP1-SETD2, KEAP1-KRAS, HGF-STK11, HGF-KRAS, BRAF-SETD2, ARID1A-TP53, ARID1A-PTEN, ARID1A-KRAS, ARID1A-HGF, ARID1A-BRAF, and BRAF-SETD2 showed significant co-occurrence, whereas EGFR-KEAP1, EGFR-STK11, EGFR-KRAS, and BRAF-EGFR exhibited mutual exclusivity across the entire cohort (p value < 0.05, Fig. 1F, Supplementary Table 2.1). In the subgroup of pN positive tumors, we observed co-occurrence for KRAS-STK11, KEAP1-STK11, and KEAP1-KRAS, whereas EGFR-STK11, BRAF-EGFR, and EGFR-KRAS were mutually exclusive (p value < 0.05, Fig. 1G, Supplementary Table 2.2). Furthermore, we identified EGFR-RBM10, KRAS-STK11, KEAP1-STK11, KEAP1-KRAS, ARID1A-TP53, ARID1A-KRAS, ARID1A-HGF, ARID1A-BRAF, and BRAF-SETD2 as significant mutual co-occurrence drivers in pN negative tumors, while EGFR-STK11, EGFR-KRAS, and BRAF-EGFR exhibited mutual exclusivity in pN negative tumors (p value < 0.05, Figure. 1H, Supplementary Table 2.3).

Our analysis, integrating alteration frequency, tumor mutation count, TMB, and MATH score, reinforces existing understanding of the genomic alterations associated with LN metastasis in LUAD. While these findings contribute to our understanding, they served as a foundation for further, more detailed investigations.

Preoperative Clinicopathologic and Genomic Features Associated with Pathologic LN Metastasis

To identify clinicopathologic and genomic features associated with pathologic LN metastasis, logistic regression analyses were performed. Univariable logistic regression analysis revealed that several preoperative clinicopathologic variables were significantly associated with pathologic LN metastasis in the entire cohort (n = 257, Supplementary Table 1, Supplementary Fig. 1A). These variables included solid versus subsolid tumor morphologic appearance on CT (p value < 0.001), tumor size on CT (p value < 0.001), histological subtype MIP/SOL (p value < 0.001), and pathological VPI and LVI (p value < 0.001). The following genomic variables were considered: EGFR mutation versus wild-type (p value = 0.087), TP53 mutation versus wild-type (p value = 0.028), RBM10 mutation versus wild-type (p value = 0.15), STK11 mutation versus wild-type (p value = 0.052), ATM mutation versus wild-type (p value = 0.055), HGF mutation versus wild-type (p value = 0.052), KEAP1 mutation versus wild-type (p value = 0.098), and PIK3CG mutation versus wild-type (p value = 0.013). The results of the multivariable regression analysis, which combined preoperative clinicopathologic and genomic features, showed that subsolid versus solid tumor morphologic appearance on CT (odds ratio [OR], 0.276; 95% confidence interval [CI], 0.085–0.89; p value = 0.032), tumor size on CT (OR, 1.036; 95% CI, 1.001–1.071; p value = 0.042), pathological LVI (OR, 3.89; 95% CI, 1.63–9.29; p value = 0.02), and the presence of PIK3CG mutation compared to wild-type (p = 0.049) were associated with pathologic LN metastasis (Supplementary Table 1, Supplementary Figure. 1B).

The study's findings reinforce existing understandings of the factors associated with LN metastasis in LUAD patients, underscoring the interplay between clinicopathologic and known genomic variables. This includes a reaffirmation of the known relevance of PIK3CG mutations in LN metastasis patterns.

The analysis of these data offered preliminary insights into the complex interplay of clinicopathologic and genomic factors in relation to LN metastasis in LUAD patients. This approach underscored the potential significance of variables such as PIK3CG mutations, although further research is necessary to fully understand their role in the pathogenesis of LN metastasis.

Oncogenic Pathways and Therapeutic Actionabilities Associated with Pathologic LN Metastasis

To further examine the association between oncogenic pathways and pathologic LN metastasis in the analysis of genomic variables above, we conducted a comparison of oncogenic signaling pathway alteration frequencies between pN positive and negative samples in the entire broad-panel NGS cohort. Among the pathways examined (Fig. 2B), the RTK/RAS pathway had the highest frequency of alterations (83%), followed by p53 (40%) and PI3K (13%). Results of the Fisher’s exact test revealed a significant increasing trend in the frequency of p53 pathway alterations from pN negative to pN positive samples (p value = 0.04, Fig. 2C). Additionally, we observed the number of pathways altered (NPA) in pN positive samples compared to pN negative samples (Fig. 2D). Nevertheless, no significant difference was found probably due to limited sample size. The COME analysis of oncogenic pathways is presented in Supplementary Fig. 2.

A further comparison of somatic mutations in genes related to genomic integrity between our cohort and the Western MSK cohort (n = 429). KRAS (12%), NF1 (1%), and BRAF (4%) of the receptor tyrosine kinase (RTK)/RAS pathway, STK11 (3%) of the phosphatidylinositol 3-kinase (PI3K) pathway, and KEAP1 (2%) of the NRF2 pathway had lower mutation frequencies, while EGFR (65%) of the RTK/RAS pathway, PIK3A (5%) of the PI3K pathway, and TP53(40%) of the p53 pathway had much higher mutation frequencies in our cohort than in Western populations (Fig. 2A). PIK3A (8%) of the PI3K pathway was detected with higher mutation frequencies in the pN positive group compared with those in the negative group (4%) in our cohort and Western cohort (6% vs 3%). The same observations were also made in BRAF of the RTK/RAS pathway in our cohort (8% vs 2%) and Western cohort (7% vs 4%). These findings indicated possible different mutation fragile sites and the significance of the RTK/RAS or PI3K pathway-related genetic mechanisms causing LUAD LN metastasis between Chinese and Western populations.

A total of 260 actionable alterations were identified using the OncoKB database, spanning 18 genes and consisting of 184 (71%) level 1, 7 (3%) level 3A, 23 (9%) level 3B, and 46 (17%) level 4 alterations (Fig. 2E-G). The RTK/RAS pathway had the highest number of actionable alterations (84% [218/260]), of which 84% (184/218) had level 1 evidence. At the sample level, actionable alterations were identified in 84% (216/257) of the 257 samples, of which 83% (180/216) had level 1 evidence. Notably, pN positive tumors showed a higher frequency tendency of level 1 actionable targets (73% vs. 60% for pN + vs. pN0, p value = 0.094). The mean number of actionable alterations per sample among subgroups did not differ significantly (1.038 vs. 1.005, all p value > 0.05). Our analyses of oncogenic pathways and therapeutic targets corroborate previously reported data, reinforcing our understanding of the molecular mechanisms associated with LN metastasis in LUAD. These investigations into pathways and therapeutic targets provided an initial framework to comprehend the molecular underpinnings of LN metastasis, suggesting the possibility of diverse genomic mutation patterns.

Identification of Cellular Metaclusters Related to LN Metastasis in TIME

To explore the spatial cellular composition of LN metastasis, we utilized mIHC to analyze primary tumor samples from 92 patients who underwent PSM in the entire cohort of broad-panel NGS. We optimized a 6-antibody panel to identify cancer cells, stromal cells, as well as innate and adaptive immune lineages with diverse functional characteristics (Fig. 3A).

We utilized a dimensionality reduction approach to classify 9 distinct cellular metaclusters (Fig. 3B, C). pN positive tumors exhibited a higher percentage of immune infiltrate (36.5%) compared to pN negative tumors (31.1%, Figure. 4D, E). Notably, the disparity was primarily driven by alterations in the macrophage (7.1% vs. 4.8%, p value < 0.05) and Treg (0.85% vs. 0.29%, p value < 0.001) populations (Figure. 3G), while no significant variations in the overall frequency of total cells were observed between different LN metastasis statuses. The prevalence of macrophages exhibited a strong correlation with the histological subtype. No significant differences were observed in the average frequency of CD4⁺T, CD8⁺T, and epithelial cellular metaclusters in total cells between pN positive and negative tumors. Our analyses involved the classification of 9 distinct cellular metaclusters in tumor samples of LUAD, emphasizing the notable contributions of macrophages and Tregs proportions in the TIME[13].

Spatial Cellular Interactions Related to LN Metastasis and Cellular Metaclusters Analyses Reveal two TIME Subtypes

In order to understand the cellular architecture and spatial organization beyond the metaclusters frequency of LN metastasis in LUAD, we analyzed the communication patterns between individual metaclusters by quantifying the spatial relationships (Fig. 4A). The distribution of the numbers of cellular connections and pairs of cell distance (µm) was presented in Supplementary Fig. 3. Furthermore, we determined the probability of interaction or avoidance behaviors between pairs of cells across tumor architectures in both pN negative and pN positive tumors (Fig. 4B). By comparing the cellular interactions within metaclusters between pN negative and pN positive tumors, macrophages were indicated at the center exhibited significant differences in spatial cell interactions with a broader range of other subclusters, including PD-L1 positive epithelial cells (Fig. 4C, box a), macrophages and T, macrophages and CD8⁺T (Fig. 4C, box b), CD4⁺T and CD8⁺T (Fig. 4C, box c). The interactions between macrophages and tumor epithelial cells in the primary tumor lesion of pN positive tumors were higher, which aligns with previous research highlighting the role of macrophages in the TIME[14, 15]. These analyses painted an overall picture of how cellular metaclusters interactions shift among LN metastasis in LUAD and exemplified how spatial relationships, rather than cell frequency alone, are important to understand patterns of spatial TIME biology.

To investigate the patterns involved in cellular metaclusters and genetic characteristics, we identified two distinct subtypes by unsupervised analysis with machine learning within the TIME of LN metastasis primary tumors: an epithelial type and a macrophage and immune-regulatory cell-enriched type (Fig. 4D). We found an association between the frequency of LN metastasis and the distribution of two clusters (p value < 0.001, Supplementary Fig. 4D), with subgroup 2 showing a higher tendency towards advanced LN stage. This novel approach[16], employing 9 metaclusters for classification, offers a new perspective in understanding the heterogeneity of the LUAD TIME related to LN metastasis.

In analyzing immune microenvironment components, significant differences were observed between CD4T cells in group 1 (mean = 0.0649) and group 2 (mean = 0.0539), with a p-value of 0.0166, suggesting varying roles of CD4T cells across these subtypes. In contrast, CD8T cells showed no significant difference between the groups (group 1 mean = 0.0600, group 2 mean = 0.0651, p value = 0.3235, Fig. 4E).

The difference in epithelial cells was particularly notable, with group 1 averaging 0.4171 and group 2 at 0.2195 (p value < 0.001), highlighting the critical role of epithelial cells in the LN metastatic TIME. Similarly, macrophages showed significant distribution differences (group 1 mean = 0.0662, group 2 mean = 0.0919, p value < 0.001, Fig. 4E), suggesting divergent functions across the two TIME subtypes.

Further insights were gained from the analysis of gene mutation frequencies across the two distictive TIME subtypes (Supplementary Fig. 4C). The ATM and PIK3CG genes exhibited the lower mutation frequency in group 1 and higher in group 2 (ATM p value = 0.013, PIK3CG p value = 0.016), potentially linked to the cell's DNA damage repair capacity and activation of immune inflammatory cells. Both BRD4 and KMT2B genes showed higher mutation frequencies in group 2 (BRD4 p value = 0.041, KMT2B p value = 0.041), indicating that these epigenetic regulators might play different roles across LUAD microenvironment subtypes. The CTNNB1 gene showed a higher mutation frequency in group 1 (p value = 0.046), associated with potential changes in the Wnt signaling pathway, impacting cell fate and proliferation.

In the subtypes of group 1, we observed co-occurrence for EGFR-TP53, KRAS-TP53 and EGFR-CTNNB1 (p value < 0.05, Supplementary Fig. 4B). Furthermore, we identified KRAS-TP53 and EGFR-PIK3CG as significant mutual co-occurrence drivers in group 2 (p value < 0.05, Supplementary Fig. 4B).

Our analyses provided initial insights into the dynamic nature of cellular metacluster interactions in LUAD, particularly in the context of LN metastasis. We suggested that spatial relationships, rather than mere cell frequency, might play a crucial role in understanding the intricacies of spatial TIME biology[17, 18].

Analysis of Distinct Cellular Neighborhood Alterations between LN Negative and Positive LUAD Patients

In the preceding section, we conducted an analysis of the variations in cellular metacluster frequencies observed between LN negative and positive primary tumors in patients with LUAD. To determine specific cellular interacting critical for the LN metastasis, we conducted cellular neighborhood (CN) and cell interaction assessments. We annotated the CNs based on the major cellular cluster within the TIME (Fig. 5A, C). First, a cellular neighborhood was designated as the ten closest neighbors surrounding the central cell. While we define CN as the centre cell, its primary neighbours within 4µm plus the secondary neighbours of the primary neighbours. Such definition is more suitable for the CN with > 10 neighbours and able to remove artificial neighbours when separate cells were in distant location. The distribution of the CN is presented in Supplementary Fig. 4A-C. Recent CN analysis framework established by Schürch et al [19] and Sheng et al [20] were adapted.

To determine various functional cellular neighborhood units, these neighborhoods were annotated based on the primary cellular clusters. Our analysis revealed distinct CN functional units, including epithelial-enriched CNs (Epithelial-CN, 2, 5, 6), macrophage and T cell-enriched CNs (MF and T-CN, 4, 7, 10), immune-suppressed-enriched CNs (Immune-suppressed-CN, 8, 9), and CNs enriched in CD8⁺ and CD4⁺ T cells (T-CN, 1). Notably, MF and T-CN exhibited coherent links with activated macrophages and macrophages and CD8⁺ T cells, whereas Immune-suppressed-CN showed links with Treg and PD-L1 positive epithelial cells (Fig. 5B,C).

We observed a distinct increase in the number of macrophage and T cell-enriched CNs (MF and T-CN, 4, 7, 10) exclusively in the LN positive LUAD patients (Fig. 5F). Similar to macrophage metacluster, CN10 also showed a counter institutive pattern that the enrichment increased specifically in LN positive group of LUAD patients (p value = 0.04). In contrast, the CN4, 7 alterations were common to both LN negative and positive group, with no metastasis-specific patterns identified in the macrophage and T cell-enriched CNs (p value > 0.05).

Similar to the changes in Treg metacluster frequency observed across LN negative and positive LUAD patients, we also noticed an increasing trend in the immune-suppressed-enriched CNs (Immune-suppressed-CN, 8, 9) formed in the LN positive primary tumors. Additionally, we found that this increasing trend in CN9 was more pronounced among the LN positive LUAD patients, which may be critical for the LN metastasis (p value = 0.002, Fig. 5E). Next, we examined the associations between the topological function units with LN stage. we applied a color-coding scheme to each mIHC image based on the CN composition, and utilized a Voronoi plot to generate a topology map for each mIHC image (Fig. 5G), employing the methodology from our previous research. The resulting topology map, encoded with CN information, accurately represented the mIHC image.However, no such trend was observed in the CN8, CNs enriched in epithelial-enriched CNs (Epithelial-CN, 2, 5, 6) and CNs enriched in CD8⁺ and CD4⁺ T cells (T-CN, 1, p value > 0.05). These results suggested that LUAD topological function units could be novel biomarkers for LN metastasis. In addition, macrophage and Tregs organized CNs with critical function in LN positive primary TIME. Patients with high level of CN9, 10 showed LN metastasis trend.

Subsequently, our analysis uncovered significant associations between the abundance of CNs and gene mutations (Supplementary Figure. 5D). Specifically, PIK3CG mutations consistently showed a higher enrichment of MF and T-CN (CN10) and T-CN (CN1), as well as a lower enrichment of epithelial-CN (CN5, CN6). Conversely, KRAS mutations were associated with a lower enrichment of MF and T-CN (CN10) and T-CN (CN1), while exhibiting a higher enrichment of epithelial-CN (CN5).

Through our investigation of spatial proteomic functional enrichment, we made several novel significant discoveries. Consistent with our previous findings of distinct macrophage-centered cellular interactions in pN positive tumors, CN enrichment analysis revealed significant heterogeneity in macrophage functionality, exhibiting distinct correlation with PIK3CG mutation. Furthermore, the enrichment of functional CNs provided a comprehensive dimension of discoveries, complementing the analysis of cell interactions. Specifically, we identified the enrichment of T cell functional units associated with PIK3CG mutation in addition to metaclusters analysis. By integrating the analysis of specific genetic mutations with multi-dimensional proteomic signatures, we generated a comprehensive spatial proteogenomic landscape of diverse LN stage LUAD primary tumor tissues.

Predicting LN Metastasis Using Machine Learning Model

In order to delve deeper into the significance of genoproteomic features in the study of LN metastasis in LUAD, and to predict the likelihood of metastasis, we investigated the potential of clinical usage. We found that the combination of NGS and spatial mIHC datasets, referred to as ImGene, is predictive of the LN stage. This led us to explore whether these data could be used to predict LN metastasis, irrespective of surgical opportunities, by applying a machine learning approach (refer to Fig. 6A and B).

Our multimodal model, which was developed using the Peking University People's Hospital (PKPH) cohort consistied of 92 patients, exhibited superior performance compared to the mIHC or genetic monomodal models. It achieved an accuracy of 0.82, an F1 score of 0.83, and an area under the curve (AUC) of 0.86 (Supplementary Fig. 6A). This model demonstrated greater accuracy in estimating the probability of N2 stage in lung cancer than both the Peking University (PKU) model-which factors in age, tumor size, central tumor location, and histological types like adenocarcinoma or adenosquamous carcinoma[21] (AUC = 0.77) and and the Fudan University model (AUC = 0.67, Supplementary Fig. 5C) for predicting LN stage[22].

In an independent validation set, the Peking University People's Hospital Thoracic Oncology Institution (PKTOI) cohort, the models achieved the following results: the ImGene integrated model had an AUC of 0.80 with an accuracy of 0.81 and an F1 score of 0.77 (Fig. 6E, Supplementary Fig. 5A), the GeneFeatures monomodel had an AUC of 0.46 (GeneFeatures monomodel), and the ImFeatures monomodels reached an AUC of 0.74 (ImFeatures monomodels, Supplementary Fig. 5C). Our algorithm effectively analyzed the immunogenomic profiles of primary tumors in LUAD to accurately predict the probability of the LN stage.

In summary, the immunogenomic model plays a crucial role in understanding the mechanisms of LUAD LN metastasis, predicting the likelihood of LN stage, and guiding pre-surgical planning as well as clinical treatment for patients who are not candidates for surgery. However, this is a complex research area that necessitates further investigation and larger-scale validation.

The identification of genomic alterations and immune cellular distribution patterns in LUAD has been pivotal in advancing targeted therapies [23–25], leading to significantly improved outcomes in metastatic disease [26–28]. However, the molecular heterogeneity of cancer [29, 30], particularly pronounced across diverse populations [31] and influenced by various survival risk factors [32, 33], underscores the necessity for a comprehensive analysis. This involves examining spatial immunogenomic patterns within different populations. Such integrative research is instrumental in gaining a deeper understanding of LN metastasis, especially in Eastern LUAD populations.

Our findings provided valuable insights into the heterogeneity of LUAD tumors with LN metastasis and highlighted the critical role of genetic alterations, cellular metaclusters and function units in modulating the LN infiltration. We observed that pN positive tumors exhibited significantly higher TMB, tumor mutation count, and intra-tumor heterogeneity compared to pN negative tumors. Additionally, our mIHC analysis identified 9 distinct cellular metaclusters within the TIME. We discerned LN stage-related TIME subtypes and functional units composed of these diverse metaclusters. Our research also highlighted distinct CN function units strongly correlated with the frequency of metaclusters, unveiling unique patterns of immune cell infiltration linked to genomic alterations in pN positive tumors relative to pN negative ones. We found that the TIME subtypes associated with pN positive tumors are characterized by a higher frequency of PIK3CG, ATM, BRD4, and KMT2B mutations, and show increased macrophage frequency alongside reduced epithelial cell infiltration. Furthermore, our study recognizes the complex interplay between genetic factors and TIME as a pivotal influence on the process of LN metastasis. Building on the genetic and TIME features, we developed a predictive model. The model effectively estimated the likelihood of LN metastasis in LUAD, marking a significant advancement in our understanding of the metastasis disease.

Our NGS findings emphasized the significance of investigating LUAD in different ethnic populations, as supported by previous research [29, 34]. Prior studies have delved into the preoperative clinical and genomic landscapes correlating with pathologic LN metastasis in early-stage LUAD, particularly stages I and II. These investigations, notably within the MSK cohort [7], elucidated significant associations between STK11 and SMARCA4 mutations and LN metastasis. In contrast, our research provided a distinct perspective, revealing new genomic dimensions in our Chinese LUAD cohort with LN metastasis. We observed a conspicuous elevation in the frequency of TP53 and PIK3CG mutations, suggesting these are pivotal genetic alterations in pN positive LUAD tumors. Our results resonated with previous literature that identified an augmented occurrence of TP53 mutations in advanced LN stage LUAD, indicating the potential implication of TP53 as a critical factor in LN metastasis, possibly mediated through p53 signaling pathways. Multiple studies have robustly associated TP53 mutations with the progression of LUAD to metastatic states, particularly affecting LN metastasis [35–37]. Functionally, TP53 mutations were instrumental in cancer proliferation, primarily by facilitating unrestrained cellular growth and division [38, 39].

Our investigation revealed a significantly higher incidence of PIK3CG alterations in pN positive tumors compared to pN negative tumors within the Chinese population cohort. This finding indicated an even greater prevalence of PIK3CG alterations than what has been reported in Western populations [40], suggesting a potential variance in the molecular underpinnings of LN metastasis across different ethnicities. Furthermore, mutations and amplifications of PIK3CG were commonly detected in a range of malignancies, including melanoma, as well as prostate, uterine, gastric, and lung cancers [41–44]. Notably, in NSCLC, the PI3K signaling pathway was significantly associated with an increased risk of metastasis [45]. In line with our findings, mutations in the PI3K-AKT-mTOR pathway, observed in various cancers including NSCLC, might be linked to the development of tumors advanced LN stages. These insights provided a robust foundation for future investigations into the role of PIK3CG in cancer lymph node metastasis.

Genetic aberrations, in synergy with the TIME, were pivotal determinants in the pathology of LN metastasis. This interplay between genetic modifications and the intricacies of TIME underscored its critical significance. Predominantly expressed in macrophages, PIK3CG displayed a marked contrast to its lower expression levels in tumor cells [46]. Investigations in PIK3CG-deficient mice showcased an augmented macrophage-mediated inflammatory response when confronted with pathogenic stimuli. Subsequent studies illuminated that tumor cells, activated the PI3K/PKB signaling cascade, culminating in enhanced expression of M2 markers and facilitating M2 macrophage polarization [47]. In congruence with these findings, our analysis delineated a comprehensive impact of PIK3CG mutations within the TIME. These mutations not only significantly altered the cellular dynamics of macrophages but also impacted their spatial CNs in both pN positive and pN negative tumor stages. Fundamentally, the extensive influence of PIK3CG mutations on the immune milieu of tumors was instrumental in elucidating the complex mechanisms underlying LN metastasis, with a distinct focus on the compositional and spatial attributes within the TIME, predominantly reflected in the activities of macrophages and Tregs.

In our study, we established a significant association of macrophages with LN metastasis, particularly evident through their interactions within spatial cellular metaclusters. Macrophages are crucial in the cancer ecosystem, facilitating communication with tumor cells and instigating inflammation [48–50], which subsequently bolsters tumor growth and invasion [51]. Additionally, their role in neovascularization, or new blood vessel formation, is essential for the metastatic spread of cancer cells, especially to the LN [52, 53]. These findings corroborate the central role of macrophages in the LN metastasis tumor microenvironment, emphasizing a spatial perspective[54]. Another study highlighted the increased expression of IL-1β in macrophages following exposure to particulate matter, indicating its necessity for EGFR-driven²³. In contrast, our research did not consistently align with EGFR-driven patterns, likely due to racial and environmental variations. Additionally, we identified that the proportion of Tregs and associated immune-suppressed-enriched CN perform distinct functions in LN metastasis. This aligns with the known role of Tregs in targeting tumor cells [55]. In summary, while prior studies have focused on cellular alterations in isolation, our research offers a comprehensive view of cellular interactions, integrating not only quantitative differences but also spatial and functional analyses.

Additionally, our investigation involved developing machine learning models that amalgamate gene alteration and mIHC data for predicting LN metastasis. The high accuracy achieved in independent external validation underscores the robustness of these models, suggesting their potential in utilizing spatial immunogenomic features to provide personalized staging recommendations for LUAD patients ineligible for surgery.

Overall, our investigation offers an in-depth analysis of the intricate relationship between genetic factors and the TIME in influencing LN metastasis in LUAD. By employing NGS and mIHC, we pinpointed intricate spatial immunogenomic patterns. These patterns yield comprehensive insights into the genomic alterations, cellular interactions, and micro-anatomical structures of TIME. Our approach in spatial proteogenomics revealed the complex interplay between LN positive tumors and their corresponding microenvironments. Such insights contribute to a deeper understanding of the factors fostering heterogeneity and pave new pathways for exploring LN metastasis in LUAD.

Limitation

Our study has several limitations that should be acknowledged. Firstly, the NGS analysis was conducted using single-region sampling of the primary tumor. This approach may not fully capture the genetic intratumoral heterogeneity inherent to LUAD, including its clonal architecture. To obtain a more comprehensive understanding of the tumor's complexity, multiple-region sampling or other techniques could be employed in future studies. Additionally, the lack of RNA sequencing data limited our ability to conduct a comprehensive analysis of the patterns and investigate the underlying mechanisms at the transcriptional level. Integrating RNA sequencing data would provide valuable insights into gene expression profiles and molecular interactions within the TIME. Future studies should consider incorporating such data to enhance the depth and robustness of the findings. Furthermore, external validation is necessary to assess the performance and generalizability of our developed model. Evaluating its predictive accuracy and reliability in independent cohorts or real-world clinical settings would provide valuable validation of its utility.

Patient cohort

This study received approval from the Institutional Review Board at Peking University People's Hospital (IRB NO.2021PHB182-001), and all patients provided written informed consent. The study enrolled patients who underwent complete resection for pathologic stage I-III LUAD and subsequent next-generation sequencing (NGS) for cancer-related genes on their primary tumor between 2018 and 2020 using 363 (HR363, Berry Oncology, Beijing, China), HR457 (Berry Oncology, Beijing, China), or RS520 (Burning Rock Biotech, Guangzhou, China). All patients underwent lobectomy with lymph node (LN) dissection to ensure the pathological assessment of LN. The exclusion criteria comprised induction therapy, wedge resection or segmentectomy, and minimally invasive adenocarcinoma (MIA).

We summarized the clinicopathologic characteristics of the broad-panel NGS cohort (n = 257), including age, sex, smoking status, tumor size, radiologic type, histological subtype, and pathological stage (adjusted based on the 8th edition of the AJCC Cancer Staging Manual). We performed propensity score matching (PSM) using a 1:1 ratio to match LN metastasis and non-metastasis cases in the NGS cohort. High inflammation versus low inflammation was the matching indicator, and the variables were age, sex, comorbidity, smoking history, histology, clinical stage. Pairs of high and low inflammation with a nearest propensityscore were matched 1:1 with a caliper width of 0.2 of standard deviation. The matching factors included age, gender, smoking, and tumor size on CT. Subsequently, we obtained a cohort of 92 patients for mIHC testing.

Tumor Genomic Analyses

We conducted target capture for the broad-panel NGS using commercial panels that included 363 (HR363, Berry Oncology, Beijing, China), 457 (HR457, Berry Oncology, Beijing, China), or 520 (RS520, Burning Rock Biotech, Guangzhou, China) cancer-related genes, covering 1.21 megabases (Mb) of the human genome. The protocols for genomic DNA extraction, targeted/whole-exome sequencing library preparation, and variant calling were described in previous publications [56].

We assessed the genomic alteration count in the broad-panel NGS cohort, defined as the sum of nonsynonymous somatic mutations, copy number variations (CNVs), and gene fusions. Additionally, we calculated the mutant-allele tumor heterogeneity (MATH) score based on the dispersion of variant allele frequency (VAF) distribution. A higher MATH score indicated a tumor with higher intratumor heterogeneity (ITH).

We applied the MutSigCV and dNdScv algorithm to infer significantly mutated genes (q < 0.1 in any caller and nonsilent mutations n ≥ 5) in the broad-panel NGS cohort (n = 257), including 53 pN positive and 204 pN negative cases. For oncogenic pathway analyses, we retained only functional alterations labeled as oncogenic, likely oncogenic, or predicted oncogenic in the OncoKB database, discarding variants of unknown significance. Therapeutic actionability information was also annotated using the OncoKB database, and each genomic alteration was stratified into one of four levels according to its clinical implication.

Multiplex Immunofluorescence (mIHC) Section Preparation

The mIHC staining was performed by Alphaxbio Biotechnology Co., Ltd. (Beijing, China), using a panel of six markers: PANCK, CD4, CD8, FOXP3, CD68, and PD-L1.

The 4-µm thick FFPE NSCLC tissue slides were subjected to deparaffinization in xylene for 30 mins followed by rehydration in absolute ethyl alcohol for 5 mins (twice), 95% ethyl alcohol for 5 mins, and 75% ethyl alcohol for 2 mins sequentially. The slides were washed three times with distilled water. Heat-induced epitope retrieval was performed using a microwave oven, and the slides were immersed in boiling EDTA buffer (PH9.0; ZLI-9069; Zsbio, Beijing, China) for 15 mins. Antibody Diluent/Block (72424205; Akoya Biosciences, DE, USA) was used for blocking before the detection of each panel, which included a total of 6 markers: PANCK, CD4, CD8, FOXP3, CD68, and PD-L1. The mIHC staining was carried out at Alphaxbio Biotechnology Co., Ltd. (Beijing, China).

Primary antibodies including FOXP3 (ab20034; Abcam, Cambridge, UK), PANCK (ab7753; Abcam, Cambridge, UK), CD8 (ab237709; Abcam, Cambridge, UK), CD68 (ab192847; Abcam, Cambridge, UK), CD4 (ab133616; Abcam, Cambridge, UK), and PDL1 (ab237726; Abcam, Cambridge, UK) were incubated at 37°C for 1 h. The slides were then incubated with Opal Polymer HRP Ms + Rb (2414515, Akoya Biosciences, USA) for 10 min at 37°C, followed by visualization using the Opal Seven-Color IHC Kit (NEL797B001KT, Akoya Biosciences, USA), with the correspondence between primary antibodies and fluorophores as XTSA520, XTSA540, XTSA570, XTSA620, XTSA650, and XTSA690. After each cycle of staining, heat-induced epitope retrieval was performed to remove all antibodies, including primary antibodies and Opal Polymer HRP Ms + Rb. Finally, the slides were counterstained with DAPI for 5 mins and enclosed in Antifade Mounting Medium (I0052; NobleRyder, Beijing, China).

mIHC Analysis and Cellular Metaclusters Identification

To analyze the tissue sections, mIHC images with high resolution in the image dataset were randomly selected as training samples to build an algorithm of tissue segmentation, cell segmentation, phenotyping tool, and positivity score using inForm Advanced Image Analysis software (inForm 2.5.0; Akoya Biosciences, USA). The algorithm was then applied in batch analysis of all the images. These procedures generated a single-cell dataset. PANCK, CD4, CD8, FOXP3, CD68, and PD-L1 were used for the subclass division of tumor cells and immune cells. In order to reduce the dimensionality of the data for visualization, Uniform manifold approximation, and projection (UMAP) was employed. Additionally, classical markers were utilized for the manual annotation of cell metaclusters.

Cellular Neighborhood (CN)Analysis

To determine the CN of each cell, we employed a methodology where windows consisting of the 20 nearest neighboring cells, based on the Euclidean distance between their X/Y coordinates[20]. These windows were then clustered based on their compositions regarding all cell metaclusters, using K-means clustering. Setting k = 10, each cell's CN was subsequently assigned according to the CN of its surrounding window[8, 57]. To validate the CN assignment, we overlaid the Voronoi allocation graphs onto the original tissue mIHC images.

Univariable and Multivariable Logistic Regression Analysis

Initially, we conducted univariate analyses regressing the LN status (pN positive or negative) on the clinicopathologic features and genomic alterations. The formula for univariable logistic regression was expressed as:

$$\:logit\left(pr\right)={\beta\:}_{0}+{\beta\:}_{1}X$$

where: pr is the probability of the LN metastasis occurring, β₀ is the intercept term, and β₁ is the coefficient for the predictor clinicopathologic or genomic variable.

The top clinicopathologic or genomic factors associated with LN metastasis in univariate analysis were included as covariates in multivariate logistic regression. The formula for multivariable logistic regression was expressed as:

The ImGene Model Construction

To predict the LN stage, we implemented each classifier using the scikit-learn (1.0.1, https://scikit-learn.org/stable/) library in Python. The coefficients were calculated as shown in Fig. 6B. We selected the top 10 coefficients from both the mIHC image features (ImFeatures) and the genomic features (GeneFeatures) to construct the models. After removing samples with unmatched ImFeatures, we obtained a dataset consisting of 83 patients with mIHC and NGS data. Among the classifiers, ImGene Support Vector Machines (SVM) demonstrated the highest performance with an AUC of 0.86, F1 Score of 0.83, and Accuracy of 0.82, surpassing Random Forest, ImFeatures SVM, GeneFeatures SVM, and Gradient Boosting.

Where, $\:\:Z=3.51+({\beta\:}_{Im}\times\:ImFeatures)+({\beta\:}_{Gene}\times\:GeneFeatures)$

$\:{\beta\:}_{Im}$ is the coefficient for the ImFeatures, $\:{\beta\:}_{Gene}$ is the coefficient for the Genefeatures (Fig. 6C).

To construct individual models based on IHC image and genomic features, SVM was implemented using three parameters: (1) C, which represents the penalty coefficient, (2) the Kernel function, and (3) Gamma. The best combination of these parameters was determined using Leave-One-Out Cross-Validation with the training dataset. The cutoff value for classification was set at the point with the highest accuracy in the validation set. In order to obtain the optimal model, logistic regression models were constructed using the predictive scores from the individual models as input features. This integration of the performance of each model was based on both the discovery and validation datasets. The probability of the LN status in a patient was calculated as follows:

$$\:Pr\left(ImGene\right)=exp\left(Z\right)/(1+exp(Z\left)\right)$$

Validation dataset

An independent external dataset comprising 61 patients treated with complete resection approved by ethics number IRB NO.2021PHB182-001 was collated. All patients provided informed consent for sample collection and all participants consented to the publication of research results. These cases were selected due to the availability of NGS and mIHC data. Clinical details for these 61 cases are summarized in Supplementary Table 4.

Statistical Analyses

The clinicopathologic features were summarized using either the median with IQR or frequency with percentage. To compare these features among the three radiological subtypes, we employed the Kruskal-Wallis test for continuous variables and the Chi-squared exact test or chi-square test for categorical variables. When it came to genomic variables, we utilized the Chi-squared test with FDR adjustment to address multiple testing. To identify the independent predictive features for LN metastasis, both univariate and multivariate negative binomial regression models were constructed, taking into account clinicopathologic and tumor genomic characteristics.

The data obtained from mIHC were imported into GraphPad software, and statistical tests were chosen based on the data distribution and variability. Two-tailed Student's t-test was used to compare two groups, while a one-way analysis of variance with Bonferroni corrected significance threshold was used for comparisons among three or more subgroups. The data are expressed as means ± SEMs (standard error of the mean). The Spearman rank correlation test was used to investigate the relationship between spatial and amount variables. The t-test and Kruskal-Wallis H test were used to compare continuous variables between two and more groups respectively. For categorical data, the Chi-squared test and Fisher's exact test were employed. For multiple clusters, the final category label was defined by combining the group information for all clusters.

All statistical analyses were performed using R 4.0.2 (R Core Team, Vienna, Austria). A p-value of < 0.05 (or FDR q < 0.1) was considered statistically significant.

Ethics approval and consent to participate

This research was approved by the Institutional Review Board at Peking University People’s Hospital (IRB NO.2021PHB182-001). The study was performed following the Declaration of Helsinki.

Availability of data and materials

All the experimental data analyzed and displayed in the present manuscript are available from the corresponding author upon reasonable request.

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

Conceptualization: Xiao Li, Yun Li, Fan Yang, Kezhong Chen, Jun Wang; Methodology: Airong Yang, Jianpeng Sheng; Formal analysis: Fanjie Meng, Hao Li, Airong Yang, Yaxing Zhao, Kaicheng Tang; Data Curation: Fanjie Meng, Airong Yang, Yaxing Zhao, Kaicheng Tang, Qing Li; Writing - Original Draft: Fanjie Meng, Hao Li, Airong Yang; Writing - Review & Editing: Ruoyi Jin, Hao Luo, Xiao Li, Peiyu Wang, Olga Chervova, Jianpeng Sheng, David Carbone; Supervision: Bin Hu, Yun Li, Fan Yang, Kezhong Chen, Jun Wang.

Funding

This work was supported by the National Natural Science Foundation of China (No.92059203, No.92259303, No.82072566, No. 82002410), Research Unit of Intelligence Diagnosis and Treatment in Early Non-small Cell Lung Cancer, Chinese Academy of Medical Sciences (2021RU002), CAMS Innovation Fund for Medical Sciences (CIFMS, 2022-I2M-C&T-B-120), Peking University Clinical Medicine Plus X - Young Scholars Project (PKU2023LCXQ008), Peking University Clinical Scientist Training Program (BMU2023PYJH011), Peking University People's Hospital Scientific Research Development Funds (RZ2022-03, RS2022-05, RDX2023-07), the Fundamental Research Funds for the Central Universities, Young Elite Scientists Sponsorship Program by CAST (2022QNRC001), Reform and Development Program of Beijing Institute of Respiratory Medicine (Ggyfz202325, Ggyfz202409), Clinical Research Incubation Project, Beijing Chao Yang Hospital, Capital Medical University (CYFH202324), Beijing Postdoctoral Research Foundation (2023-ZZ-042).

Acknowledgments

The authors wish to acknowledge the contribution of Yun Wang in the laboratory of Peking University People’s Hospital for technical assistance, and Figdraw (https://www.figdraw.com/) for flow chart generation.

Heiden BT, Eaton DB Jr, Chang SH, et al. Assessment of Updated Commission on Cancer Guidelines for Intraoperative Lymph Node Sampling in Early Stage NSCLC. J Thorac Oncol. 2022;17(11):1287–96.
Rami-Porta R, Asamura H, Travis WD, Rusch VW. Lung cancer - major changes in the American Joint Committee on Cancer eighth edition cancer staging manual. CA Cancer J Clin. 2017;67(2):138–55.
Lim E, McElnay PJ, Rocco G, et al. Invasive mediastinal staging is irrelevant for PET/CT positive N2 lung cancer if the primary tumour and ipsilateral lymph nodes are resectable. Lancet Respir Med. 2015;3(9):e32–3233.
Nguyen B, Fong C, Luthra A, et al. Genomic characterization of metastatic patterns from prospective clinical sequencing of 25,000 patients. Cell. 2022;185(3):563–e7511.
Patten DK, Corleone G, Győrffy B, et al. Enhancer mapping uncovers phenotypic heterogeneity and evolution in patients with luminal breast cancer. Nat Med. 2018;24(9):1469–80.
Siravegna G, Mussolin B, Buscarino M, et al. Clonal evolution and resistance to EGFR blockade in the blood of colorectal cancer patients. Nat Med. 2015;21(7):795–801.
Caso R, Connolly JG, Zhou J, et al. Preoperative clinical and tumor genomic features associated with pathologic lymph node metastasis in clinical stage I and II lung adenocarcinoma. NPJ Precis Oncol. 2021;5(1):70.
Ji Y, Sun D, Zhao Y. A high-throughput mass cytometry barcoding platform recapitulating the immune features for HCC detection. Nano Today. 2023:52.
Sorin M, Rezanejad M, Karimi E, et al. Single-cell spatial landscapes of the lung tumour immune microenvironment. Nature. 2023;614(7948):548–54.
Che G, Yin J, Wang W, et al. Circumventing drug resistance in gastric cancer: A spatial multi-omics exploration of chemo and immuno-therapeutic response dynamics. Drug Resist Updat. 2024;74:101080.
Seferbekova Z, Lomakin A, Yates LR, Gerstung M. Spatial biology of cancer evolution. Nat Rev Genet. 2023;24(5):295–313.
Martínez-Ruiz C, Black J, Puttick C, et al. Genomic-transcriptomic evolution in lung cancer and metastasis. Nature. 2023;616(7957):543–52.
Zhang J, Song J, Tang S, et al. Multi-omics analysis reveals the chemoresistance mechanism of proliferating tissue-resident macrophages in PDAC via metabolic adaptation. Cell Rep. 2023;42(6):112620.
Christofides A, Strauss L, Yeo A, Cao C, Charest A, Boussiotis VA. The complex role of tumor-infiltrating macrophages. Nat Immunol. 2022;23(8):1148–56.
Sheng J, Ruedl C, Karjalainen K. Most Tissue-Resident Macrophages Except Microglia Are Derived from Fetal Hematopoietic Stem Cells. Immunity. 2015;43(2):382–93.
Shao W, Zuo Y, Shi Y, et al. Characterizing the Survival-Associated Interactions Between Tumor-Infiltrating Lymphocytes and Tumors From Pathological Images and Multi-Omics Data. IEEE Trans Med Imaging. 2023;42(10):3025–35.
Sheng J, Chen Q, Soncin I, Ng SL, Karjalainen K, Ruedl C. A Discrete Subset of Monocyte-Derived Cells among Typical Conventional Type 2 Dendritic Cells Can Efficiently Cross-Present. Cell Rep. 2017;21(5):1203–14.
Shao W, Liu J, Zuo Y, Qi S, Hong H, Sheng J, Zhu Q, Zhang D. FAM3L: Feature-Aware Multi-Modal Metric Learning for Integrative Survival Analysis of Human Cancers. IEEE Trans Med Imaging. 2023;42(9):2552–65.
Schürch CM, Bhate SS, Barlow GL, et al. Coordinated Cellular Neighborhoods Orchestrate Antitumoral Immunity at the Colorectal Cancer Invasive Front. Cell. 2020;182(5):1341–e5919.
Sheng J, Zhang J, Wang L, et al. Topological analysis of hepatocellular carcinoma tumour microenvironment based on imaging mass cytometry reveals cellular neighbourhood regulated reversely by macrophages with different ontogeny. Gut. 2022;71(6):1176–91.
Chen K, Yang F, Jiang G, Li J, Wang J. Development and validation of a clinical prediction model for N2 lymph node metastasis in non-small cell lung cancer. Ann Thorac Surg. 2013;96(5):1761–8.
Verdial FC, Madtes DK, Hwang B, Mulligan MS, Odem-Davis K, Waworuntu R, Wood DE, Farjah F. Prediction Model for Nodal Disease Among Patients With Non-Small Cell Lung Cancer. Ann Thorac Surg. 2019;107(6):1600–6.
Zhang C, Zhang J, Xu FP, et al. Genomic Landscape and Immune Microenvironment Features of Preinvasive and Early Invasive Lung Adenocarcinoma. J Thorac Oncol. 2019;14(11):1912–23.
Kuzbari Z, Bandlamudi C, Loveday C, et al. Germline-focused analysis of tumour-detected variants in 49,264 cancer patients: ESMO Precision Medicine Working Group recommendations. Ann Oncol. 2023;34(3):215–27.
Parra ER, Zhang J, Jiang M, et al. Immune cellular patterns of distribution affect outcomes of patients with non-small cell lung cancer. Nat Commun. 2023;14(1):2364.
Solomon BJ, Bauer TM, Mok T, et al. Efficacy and safety of first-line lorlatinib versus crizotinib in patients with advanced, ALK-positive non-small-cell lung cancer: updated analysis of data from the phase 3, randomised, open-label CROWN study. Lancet Respir Med. 2023;11(4):354–66.
Drilon A, Oxnard GR, Tan D, et al. Efficacy of Selpercatinib in RET Fusion-Positive Non-Small-Cell Lung Cancer. N Engl J Med. 2020;383(9):813–24.
Le X, Cornelissen R, Garassino M, et al. Poziotinib in Non-Small-Cell Lung Cancer Harboring HER2 Exon 20 Insertion Mutations After Prior Therapies: ZENITH20-2 Trial. J Clin Oncol. 2022;40(7):710–8.
Chen J, Yang H, Teo A, et al. Genomic landscape of lung adenocarcinoma in East Asians. Nat Genet. 2020;52(2):177–86.
Yang W, Zhao Y, Ge Q, et al. Genetic mutation and tumor microbiota determine heterogenicity of tumor immune signature: Evidence from gastric and colorectal synchronous cancers. Front Immunol. 2022;13:947080.
Rodon J, Soria JC, Berger R, et al. Genomic and transcriptomic profiling expands precision cancer medicine: the WINTHER trial. Nat Med. 2019;25(5):751–8.
Hill W, Lim EL, Weeden CE, et al. Lung adenocarcinoma promotion by air pollutants. Nature. 2023;616(7955):159–67.
Sitas F, Egger S, Bradshaw D, Groenewald P, Laubscher R, Kielkowski D, Peto R. Differences among the coloured, white, black, and other South African populations in smoking-attributed mortality at ages 35–74 years: a case-control study of 481,640 deaths. Lancet. 2013;382(9893):685–93.
Zhao Y, Murciano-Goroff YR, Xue JY, et al. Diverse alterations associated with resistance to KRAS(G12C) inhibition. Nature. 2021;599(7886):679–83.
Lengel HB, Mastrogiacomo B, Connolly JG et al. Genomic mapping of metastatic organotropism in lung adenocarcinoma. Cancer Cell. 2023.
Powell E, Piwnica-Worms D, Piwnica-Worms H. Contribution of p53 to metastasis. Cancer Discov. 2014;4(4):405–14.
Hu J, Cao J, Topatana W, et al. Targeting mutant p53 for cancer therapy: direct and indirect strategies. J Hematol Oncol. 2021;14(1):157.
Muller PA, Caswell PT, Doyle B, et al. Mutant p53 drives invasion by promoting integrin recycling. Cell. 2009;139(7):1327–41.
Arjonen A, Kaukonen R, Mattila E, et al. Mutant p53-associated myosin-X upregulation promotes breast cancer invasion and metastasis. J Clin Invest. 2014;124(3):1069–82.
Grosse C, Soltermann A, Rechsteiner M, Grosse A. Oncogenic driver mutations in Swiss never smoker patients with lung adenocarcinoma and correlation with clinicopathologic characteristics and outcome. PLoS ONE. 2019;14(8):e0220691.
Shorning BY, Dass MS, Smalley MJ, Pearson HB. The PI3K-AKT-mTOR Pathway and Prostate Cancer: At the Crossroads of AR, MAPK, and WNT Signaling. Int J Mol Sci. 2020;21(12):4507.
Comprehensive genomic characterization. of squamous cell lung cancers. Nature. 2012;489(7417):519–25.
Comprehensive molecular characterization of gastric adenocarcinoma. Nature. 2014;513(7517):202–9.
Genomic Classification of Cutaneous Melanoma. Cell. 2015;161(7):1681–96.
Fruman DA, Chiu H, Hopkins BD, Bagrodia S, Cantley LC, Abraham RT. The PI3K Pathway in Human Disease. Cell. 2017;170(4):605–35.
Kaneda MM, Messer KS, Ralainirina N, et al. PI3Kγ is a molecular switch that controls immune suppression. Nature. 2016;539(7629):437–42.
Wang X, Luo G, Zhang K, et al. Hypoxic Tumor-Derived Exosomal miR-301a Mediates M2 Macrophage Polarization via PTEN/PI3Kγ to Promote Pancreatic Cancer Metastasis. Cancer Res. 2018;78(16):4586–98.
Lewis CE, Pollard JW. Distinct role of macrophages in different tumor microenvironments. Cancer Res. 2006;66(2):605–12.
Villanueva MT. Cancer immunotherapy: Macrophages steal the show. Nat Rev Drug Discov. 2017;16(7):455.
Kloosterman DJ, Akkari L. Macrophages at the interface of the co-evolving cancer ecosystem. Cell. 2023;186(8):1627–51.
Goossens P, Rodriguez-Vita J, Etzerodt A, et al. Membrane Cholesterol Efflux Drives Tumor-Associated Macrophage Reprogramming and Tumor Progression. Cell Metab. 2019;29(6):1376–e894.
Ginhoux F, Jung S. Monocytes and macrophages: developmental pathways and tissue homeostasis. Nat Rev Immunol. 2014;14(6):392–404.
Bieniasz-Krzywiec P, Martín-Pérez R, Ehling M, et al. Podoplanin-Expressing Macrophages Promote Lymphangiogenesis and Lymphoinvasion in Breast Cancer. Cell Metab. 2019;30(5):917–e3610.
Du J, Zhang J, Wang L, et al. Selective oxidative protection leads to tissue topological changes orchestrated by macrophage during ulcerative colitis. Nat Commun. 2023;14(1):3675.
Kumagai S, Togashi Y, Sakai C, et al. An Oncogenic Alteration Creates a Microenvironment that Promotes Tumor Progression by Conferring a Metabolic Advantage to Regulatory T Cells. Immunity. 2020;53(1):187–e2038.
Chen K, Yang A, Carbone DP, et al. Spatiotemporal genomic analysis reveals distinct molecular features in recurrent stage I non-small cell lung cancers. Cell Rep. 2022;40(2):111047.
Bao X, Li Q, Chen D, et al. A multiomics analysis-assisted deep learning model identifies a macrophage-oriented module as a potential therapeutic target in colorectal cancer. Cell Rep Med. 2024;5(2):101399.

No competing interests reported.

Download PDF

Editorial decision: Revision requested
24 Aug, 2024
Reviews received at journal
11 Aug, 2024
Reviews received at journal
07 Aug, 2024
Reviewers agreed at journal
06 Aug, 2024
Reviewers agreed at journal
02 Aug, 2024
Reviewers agreed at journal
01 Aug, 2024
Reviewers invited by journal
01 Aug, 2024
Editor assigned by journal
30 Jul, 2024
Submission checks completed at journal
30 Jul, 2024
First submitted to journal
30 Jul, 2024

You are reading this latest preprint version

Spatial Immunogenomic Patterns Associated with Lymph Node Metastasis in Lung Adenocarcinoma

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Figures

Background

Results

Genomic Features Associated with Pathologic LN Metastasis in Eastern Asian LUAD

Preoperative Clinicopathologic and Genomic Features Associated with Pathologic LN Metastasis

Oncogenic Pathways and Therapeutic Actionabilities Associated with Pathologic LN Metastasis

Identification of Cellular Metaclusters Related to LN Metastasis in TIME

Spatial Cellular Interactions Related to LN Metastasis and Cellular Metaclusters Analyses Reveal two TIME Subtypes

Analysis of Distinct Cellular Neighborhood Alterations between LN Negative and Positive LUAD Patients

Predicting LN Metastasis Using Machine Learning Model

Discussion

Limitation

Methods

Patient cohort

Tumor Genomic Analyses

Multiplex Immunofluorescence (mIHC) Section Preparation

mIHC Analysis and Cellular Metaclusters Identification

Cellular Neighborhood (CN)Analysis

Univariable and Multivariable Logistic Regression Analysis

The ImGene Model Construction

Validation dataset

Statistical Analyses

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1