Abstract

   The inconsistent immunotherapy response among lesions in patients with
   multiple primary lung cancer (MPLC) remains poorly understood,
   presenting a significant challenge for effective treatment. In this
   study, we conducted a comprehensive multiomics analysis of all lesions
   from a patient with MPLC who exhibited varied responses to neoadjuvant
   chemoimmunotherapy. Further verification was conducted through external
   single-cell data and multiplex immunohistochemistry of three cases of
   MPLC.Notably, tertiary lymphoid structures (TLSs) were observed across
   all nodules, regardless of response, suggesting possible TLS impairment
   in non-responsive nodules. Cell neighborhood (CN) analysis revealed
   that type II alveolar epithelial cell (AT2) cell–positive CNs were
   prevalent in non-responsive nodules, while AT2-negative CNs appeared in
   responsive nodules, strongly associating AT2 cell presence with a
   reduced therapeutic response. Spatial colocalization analysis further
   showed that AT2 cells surrounding TLSs upregulated immunosuppressive
   markers on B cells within TLSs. The mechanism of this suppressive
   effect was further unveiled that macrophage migration inhibitory factor
   (MIF), secreted by AT2 cells, binds to sialic acid acetylesterase
   (SIAE) receptors on B cells by single-cell RNA sequencing analysis,
   which were validated in four additional non-responsive nodules from
   three other patients with MPLC. Multiomics analysis revealed AT2 cells
   exert immunosuppressive effects by inhibiting B cells within TLS
   through MIF-SIAE signaling axis in patients with MPLC. These findings
   offered new perspectives for tailored immunotherapy for patients with
   MPLC.

   Keywords: Lung Cancer, Neoadjuvant, Immunotherapy

Introduction

   Multiple primary lung cancers (MPLCs) refer to two or more independent
   primary lung cancer lesions that occur simultaneously or at different
   times in the lungs of the same patient. According to the diagnostic
   interval time, MPLCs can be divided into two categories: synchronous
   (interval ≤6 months) and metachronous (interval >6 months).[49]^1 With
   the increasing incidence of lung cancer and growing use of early chest
   CT screening, the proportion of MPLCs has significantly increased, from
   less than 5% at the beginning of this century to nearly 15%, and is
   still on the rise, posing a significant challenge for clinical
   discrimination and treatment.[50]^2 3 The etiology of MPLC involves the
   interaction of multiple factors, including regional carcinogenesis
   (such as smoking and air pollution), genetic susceptibility and
   age-related factors, etc.[51]^4

   Patients with MPLC are usually diagnosed at an early stage, often
   presenting with concurrently multiple concurrent subsolid nodules
   (SSNs), making surgery the preferred treatment for operable patients.
   However, since multiple SSNs cannot always be resected simultaneously,
   other treatment strategies such as radiotherapy, targeted therapy and
   immunotherapy can be selected according to tumor characteristics.[52]^5
   With the application of neoadjuvant immunotherapy for early-stage
   Non-Small Cell Lung Cancer (NSCLC), some studies have begun to evaluate
   the feasibility of neoadjuvant chemoimmunotherapy in MPLC.[53]^6 7
   Unfortunately, Immune Checkpoint Inhibitors (ICIs) do not consistently
   target all lesions in patients with MPLC, and the reasons for this
   variability remain unclear.[54]^8 The MPLC lesions with heterogeneous
   immune responses provide an excellent opportunity to explore the
   response mechanism of neoadjuvant chemoimmunotherapy since all lesions
   share homogeneous pathogenic backgrounds.

   Here, we conducted a comprehensive spatial multiomics analysis,
   particularly using imaging mass cytometry (IMC), for a patient with
   MPLC receiving neoadjuvant chemoimmunotherapy. Our study uncovered
   critical spatial cell interactions that influence immunotherapy
   efficacy, offering valuable insights into potential targets for future
   immunotherapy strategies.

Results

   An adult patient was admitted to our hospital after a positron emission
   tomography scan that revealed one main solid nodule and three SSNs with
   enlarged right lower paratracheal lymph nodes (R4LN), which was further
   confirmed as lung adenocarcinoma by endobronchial needle aspiration
   biopsy ([55]online supplemental figure S1A). Next-generation sequencing
   of the R4LN specimen based on a panel of 520 genes revealed 14
   mutations, including common oncogenic mutations in EGFR and TP53.
   Notably, the patient had two types of EGFR mutations, including
   21p.L858R and 20p.Q791L ([56]online supplemental table S1). Although
   studies have found that approximately 30%–40% of MPLCs have EGFR
   mutations, the EGFR mutations observed in this case represented a
   patient-specific genomic event rather than a universal feature of MPLC
   pathogenesis.[57]9,[58]11 Immunohistochemistry revealed 95% Programmed
   death-ligand 1 (PD-L1) expression, suggesting a high likelihood of
   benefit from ICI therapy. After three cycles of neoadjuvant
   chemoimmunotherapy (pemetrexed, carboplatin, pembrolizumab), chest CT
   showed significant shrinkage of solid nodule and R4LN, while three SSNs
   remained stable ([59]online supplemental figure S1B). A surgical
   resection of the right upper lobe and mediastinal lymph nodes was
   performed. Pathological examination revealed a pathological complete
   response (pCR) of solid nodule as primary tumor (PT) and all the lymph
   nodes. The additional three SSNs had 100% residual viable tumor cells
   and were diagnosed as two minimally invasive adenocarcinomas, labeled
   SSN1 and SSN2, and one atypical adenomatous hyperplasia, labeled SSN3.
   Notably, tertiary lymphoid structures (TLSs) were discerned in all
   lesions via histological examination ([60]online supplemental figure
   S1B), suggesting a potential therapeutic response.[61]^12 However, such
   responses did not yield significant reductions of SSNs. Variations in
   treatment efficacy among MPLCs underscore the need to investigate the
   mechanisms that compromise TLS function.

   To explore different features of genomic and tumor immune
   microenvironments (TIME) among MPLC lesions, we performed whole exome
   sequencing (WES) and bulk RNA sequencing on each resected sample. Each
   lesion presents with a similar number of scarce mutations, aligns with
   findings of reduction in sample mutation frequency post-ICI
   treatment[62]^13 14 ([63]online supplemental figure S1C). In all tumor
   lesions, we observed the residual mammalian Ste20-like kinases 1 gene
   variations on the Hippo signaling pathway, suggesting that it may be
   related to tumor progression and drug resistance.[64]^15 16 Arm-level
   somatic copy number alterations indicated non-responsive SSNs,
   especially SSN2, have higher genomic instability, which may contribute
   to reduced immune efficacy. Consistent with the H&E staining results,
   we found that TLS signature scores were high in all samples regardless
   of immune response ([65]online supplemental figure S1D). PT exhibited
   an immune-desert phenotype with significant fibroblast enrichment,
   indicating stromal remodeling after tumor elimination. In contrast,
   R4LN demonstrated robust immune infiltration, with substantial
   accumulation of lymphocytes including T cells, CD8+T cells, and B
   cells. This difference in immune infiltration underscores fundamental
   differences in the response mechanism between tumor lesions and the
   lymph nodes. All three non-responding SSNs were enriched in monocyte
   lineage cells, neutrophils and endothelial cells. It is worth noting
   that compared with SSN2, SSN1 and SSN3 contain upregulated cytotoxic
   tumor-infiltrating lymphocytes, suggesting functional heterogeneity
   related to the maturity of TLS across lesions. More genes were
   significantly upregulated in non-responsive lesions, especially
   interleukin-1 receptor-like 1 (IL1RL1) and vascular endothelial growth
   (VEGFA), contributing to the TIME remodeling ([66]online supplemental
   figure S1E). Pathway enrichment analysis showed that B cells and T
   cells are activated through signaling of antigen recognition to enhance
   immunity in responsive PT. In contrast, non-responsive SSNs were
   significantly involved in cell proliferation and epithelium
   differentiation, which could partially explain the limited response in
   SSNs ([67]online supplemental figure S1F). Those highly heterogeneous
   transcriptional features among MPLC lesions suggest a need for further
   in-depth studies of the TIME features.

   To spatially characterize the TIME landscape of MPLC lesions, we
   applied a 38-plex antibody panel IMC to all lesions ([68]online
   supplemental figure S2). For each lesion, we selected three to four
   regions of interest (ROIs) (total ROIs=20) to fully display their TIME
   features. We identified 2,00,139 cells across 17 populations, with T
   cells further divided into eight functional subgroups ([69]online
   supplemental figure S3A, B). The distribution and absolute density of
   cell cluster types (CTs) across all ROIs were analyzed, with
   tissue-resident macrophages being the most abundant ([70]online
   supplemental figure S3C). Comparative analysis of the cell populations
   ([71]online supplemental figure S4 and [72]online supplemental figure
   S5) across all three SSNs revealed no significant differences,
   indicating a consistent immunological landscape among them. Therefore,
   the SSNs were grouped together for subsequent analyses as
   representative non-responsive lesions. Non-responsive SSNs exhibited
   higher levels of AT2 (type II alveolar epithelial cell), CD1c+dendritic
   cells, CD11b+granulocytes, CD66b+granulocytes, and natural killer T
   cells (NKTs) compared with responsive PT ([73]online supplemental
   figure S3D and [74]online supplemental figure S4). In T-cell subsets,
   Regulatory T cells (Tregs) were enriched in responsive PT, while naive
   T cells dominated non-responsive SSNs. However, activated T-cell levels
   showed no significant difference, suggesting they may not be the key
   determinant of immunotherapy efficacy ([75]online supplemental figure
   S3E). Overall, the spatial cellular composition of non-responsive SSNs
   demonstrated an immune-inactive state.

   To elucidate the effect of cell interactions based on the spatial
   proximity of different metaclusters in TIME, we conducted regional
   cellular neighborhoods (CNs) analysis. Here, CNs were defined as the 20
   closest neighboring cells surrounding each central cell ([76]figure
   1A). We used a previously published color-coding scheme to generate a
   topology map for each IMC image[77]^17 to better illustrate the
   topological CN composition ([78]figure 1B, C).

Figure 1. Functional cell neighborhoods analysis and cellular interaction
profiles in different lesions. (A) Schematic of spatial CN and cell–cell talk
analysis. (B) CN model diagram and corresponding H&E and imaging mass
cytometry images. (C) Heatmap of 20 CNs and CN annotation and cell
composition (left) and pattern maps of representative CN2 and CN8 (right).
(D) CN frequency comparison and significant difference between different
groups. (E) Heatmap depicting significant pairwise cellular interaction (red)
or avoidance (blue) between differential metaclusters in responsive and
non-responsive lesions. A one-way analysis of variance test was used for
statistical analysis across clinical variables. AT2 cell, type II alveolar
epithelial cell; CN, cell neighborhood; DC, dendritic cell; HEV, high
endothelial venules; PT, primary tumor; R4LN, right lower paratracheal lymph
nodes; SSN, subsolid nodule; Treg, Regulatory T cell.

   [79]Figure 1
   [80]Open in a new tab

   The AT2-positive niche (CN2) and AT2-negative CN (CN8) exhibited
   opposing patterns across lesions, with non-responsive SSNs showing
   higher CN2 and lower CN8, while responsive PT displayed the reverse
   trend ([81]figure 1D, [82]online supplemental figure S5). Furthermore,
   the AT2-enriched CN (CN20) was also prevalent in non-responsive SSNs,
   suggesting an immunosuppressive role of AT2 cells in the TIME through
   interactions with surrounding cells, ultimately reducing immunotherapy
   efficacy ([83]figure 1D). Intercellular communication analysis revealed
   distinct spatial avoidance patterns between AT2 epithelial cells and
   TLS-associated components, including B-cell lineages (B/mature B cells)
   and high endothelial venules (HEVs). This avoidance effect is
   particularly pronounced in non-responsive SSNs, with a statistically
   significant difference in the TLS avoidance effect of AT2 ([84]figure
   1E, indicated by the yellow box). This pattern of spatial coexistence
   yet functional antagonism suggests a microenvironmental exclusion
   mechanism whereby AT2-dominated niches may actively impede TLS
   maturation in SSN lesions. Taking together, spatial cell neighborhood
   and interaction analysis demonstrated that AT2 cells exert a stronger
   inhibitory effect on TLS in non-responsive SSNs, further elucidating
   the mechanism of TLS dysfunction.

   To assess the link between TLS dysfunction in SSNs and AT2 enrichment,
   we analyzed cell clusters and immune checkpoints expression inside and
   outside TLSs ([85]figure 2A). A “TLS-patch” was defined as ≥10 CD20+ B
   cells with an intercell distance of ≤20 µm according to the established
   IMC image analysis framework.[86]^18 There was no significant
   difference in the size and number of TLS patches between different
   lesions, suggesting that TLSs may operate through function regulation
   rather than quantitative differences ([87]figure 2B). Mature B cells
   were the predominant cell type within TLSs; however, their numbers were
   similar across lesions, indicating that mature B-cell presence alone
   does not determine immune efficacy. AT2 cells were primarily located
   outside TLS and were more numerous in SSNs with poor immunotherapy
   response ([88]figure 2C). Moreover, multiple inhibitory immune
   checkpoints were highly expressed within TLSs in non-responsive SSNs,
   and were significantly positively correlated with AT2 cell frequency.
   Spatial colocalization found that AT2 outside the patch encircled the
   TLSs and closely associated with these immune checkpoint markers
   ([89]figure 2D). In summary, “TLS-patch” analysis confirmed the spatial
   proximity of AT2 cells and mature B cells, along with the higher
   expression of immune checkpoints within TLSs. This demonstrated that
   AT2 cells could upregulate immunosuppressive markers in B cells within
   TLSs among non-responsive SSNs, reducing their responsiveness to
   immunotherapy.

Figure 2. Multiomics analysis revealed AT2 cells drive immunosuppression by
inhibiting B cells within TLS. (A) Definitions inside and outside TLS,
including model diagram, H&E diagrams, patch diagram, and IMC diagram. (B)
The number and size of TLS among different lesions. (C) Difference in AT2 and
B cells frequency inside and outside TLS. (D) Correlation between the
expression of immune checkpoints in mature B cells and AT2 cell frequency in
TLS (up) and spatial location of individual immune checkpoints in mature B
cells and AT2 cells on IMC image (bottom). (E) Correlation heatmap of
mutation and differential gene correlation with related cluster type and
cellular neighborhood. (F) UMAP dimensionality reduction cluster and the
proportion of cell clusters in each lesion. (G) Prediction of possible
ligands and target genes from AT2 to B cells in non-responsive nodules. (H)
Representative multiplex immunohistochemistry staining for B (green), AT2
(rose red), MIF (yellow), SIAE (light blue), and DAPI (blue) verifies the
interactive pairs. Scale bar, 50 µm. AT2 cell, type II alveolar epithelial
cell; DAPI, 4′,6-diamidino-2-phenylindole; IMC, imaging mass cytometry;
LAMP3, lysosomal-associated membrane protein 3; MIF, macrophage migration
inhibitory factor; PT, primary tumor; R4LN, right lower paratracheal lymph
nodes; SIAE, sialic acid acetylesterase; SSN, subsolid nodule; TLS, tertiary
lymphoid structure; UMAP, Uniform Manifold Approximation and Projection.

   [90]Figure 2
   [91]Open in a new tab

   Multiomics analysis integrating IMC and bulk RNA sequencing showed that
   mature B cells correlated positively with AT2-associated CNs (CN2,
   CN20) and negatively with CN8, supporting AT2-B cell interactions.
   Additionally, upregulated genes in non-responsive nodules, such as
   IL1RL1 and VEGFA, were strongly linked to AT2 frequency, suggesting
   AT2-driven TIME remodeling impacts TLS function ([92]figure 2E).

   In order to further clarify the mechanism of AT2-B cell interactions,
   we leveraged previously published single-cell RNA sequencing data
   ([93]GSE146100) from an independent female patient with MPLC with three
   pulmonary nodules.[94]^6 One nodule in the left upper lobe was a solid
   lesion, while the rest were SSNs. After three cycles of pembrolizumab
   treatment, the solid nodules showed significant reduction on CT, while
   the other two SSNs remained stable or slightly enlarged. The proportion
   of AT2 and B cells in non-responsive lesions is higher than that in
   responsive lesions, which is consistent with our findings ([95]figure
   2F). Ligand-receptor prediction analysis indicated that AT2 cells may
   exert an immunosuppressive effect on B cells through the macrophage
   migration inhibitory factor (MIF)[96]^19 ligand and sialic acid
   acetylesterase (SIAE)[97]^20 receptor in non-responsive SSNs
   ([98]figure 2G). The mechanism of TLS dysfunction in non-responsive
   SSNs is likely to be that the surrounding AT2 cells secrete MIF, which
   binds to the SIAE receptor on B cells to inhibit antigen receptor
   signaling, thereby dampening the antitumor immune response. This
   interaction was further validated by multiplex immunohistochemistry
   (mIHC) staining of four non-reactive SSNS surgically resected from
   another three patients with MPLC after neoadjuvant chemoimmunotherapy
   ([99]figure 2H, [100]online supplemental figure S6, [101]online
   supplemental table S2, [102]online supplemental table S3). All
   unresponsive nodules exhibited a considerable degree of colocalization
   of AT2-secreted MIFs and SIAE receptors on TLSs ([103]online
   supplemental table S4), supporting an immunosuppressive role of the
   MIF-SIAE signaling axis in these lesions.

Discussion

   Although a few studies have explored the response mechanisms of
   neoadjuvant chemoimmunotherapy for lung cancer, none have addressed the
   heterogenous spatial TIME features of MPLC. For the first time, we
   uncovered distinct topological and cellular crosstalk patterns within
   spatial TIME across MPLC lesions with varying immune responses, using a
   single-cell spatial multiomics approach. We successfully used highly
   multiplexed IMC to demonstrate the crucial CTs and CNs patterns that
   drive immune responses. Furthermore, we apply a novel “TLS-patch”
   analysis paradigm to elucidate the impact of encircling AT2 cells on
   TLS function, which were further validated by single-cell RNA (scRNA)
   sequencing and mIHC of independent patients.

   Generally, the presence of TLS in solid tumors is associated with a
   better prognosis and clinical outcome after immunotherapy. However,
   some studies have not found a positive impact of TLS on immunotherapy
   efficacy, particularly in patients with MPLC.[104]^5 Consistent with
   their findings, our study found CD20-labeled TLSs in all four lesions
   without significant differences between different responding lesions.
   We suspected that immunosuppressive components within TIME directly
   inhibit TLS function in non-responsive SSNs. Notably, AT2 cells were
   significantly enriched in non-responsive SSNs and closely related to
   the spatial location of TLS, suggesting that AT2 cells perform
   immunosuppressive functions by influencing TLS. AT2-free CN was the
   only cell neighborhood strongly associated with immune-responsive PT,
   further demonstrating that the inhibitory function of AT2 cells is a
   key factor in poor immune response. Recently, AT2 cells have been
   proved to be the origin of lung adenocarcinoma, promoting tumor
   progression through interactions with immune cell infiltration in the
   TIME.[105]^21 Studies have shown that the AT2 major histocompatibility
   complex type II tumor antigen can activate cytotoxic T cells to kill
   tumors.[106]^22 In unresponsive SSNs, although AT2 cells were
   significantly enriched, T cells were all in an immature state and could
   not be activated to exert killing effects. These findings highlight AT2
   cells as a novel therapeutic target to improve the efficacy of
   neoadjuvant immunotherapy.

   Further “TLS-patch” analysis showed that AT2 cells encircled TLSs exert
   immunosuppressive effects by upregulating immune checkpoints expression
   of B cell within TLSs, mediated through MIF ligands in non-responsive
   SSNs. Studies have shown that the expression of MIF is related to the
   occurrence of MPLC, and the upregulation of MIF in MPLCs leads to
   morphological changes of AT2 cells, promoting their proliferation,
   migration and invasion.[107]^19 Our study identified the role of AT2
   cells in inhibiting B-cell immunity in TLS through the MIF-SIAE axis in
   non-responsive SSNs and validated by mIHC in three independent patients
   with MPLC. We acknowledge that the relatively limited sample size in
   our study may affect the broader applicability of our conclusions. In
   the future, we will further conduct multicenter cooperation or expand
   the sample size to verify the findings, with the expectation of
   translating and applying them to clinical practice.

   Our limitations primarily stem from a single case study with only one
   responsive tumor, though validated with three additional patients with
   MPLC. Due to differences in the immune microenvironments of the PT and
   the lymph node, these two reactive lesions were analyzed separately. We
   found that, compared with PT, lymph nodes (LN) significantly enriched
   NKT cells and HEV enriched CN (CN17), which also indicated the
   differences in the immune microenvironment between PT and LN. The
   immune response mechanism of LN remains to be further explored.
   Besides, WES analysis did not detect valuable mutations including EGFR
   since the patient reached pCR after neoadjuvant therapy.

Conclusion

   In summary, through innovative comprehensive multiomics studies on
   individual cases, we characterized the immune microenvironment in
   poorly responsive nodules of MPLC and identified that AT2 cells can
   impair TLS functionality by suppressing B-cell activity. This discovery
   offers new insights into overcoming poor immunotherapeutic responses in
   MPLCs, potentially leading to improved outcomes for patients with lung
   cancer.

Methods

Study samples and pathological response evaluation

   This study received approval from the Institutional Review Board at
   Peking University People’s Hospital (IRB NO.2021PHB182-001). We
   describe patient information according to the Case Report (CARE)
   checklist for case reports ([108]online supplemental file 1). In
   addition to the main case of MPLCs described above, we also included
   three patients with MPLC who received neoadjuvant chemoimmunotherapy
   and lung resection ([109]online supplemental table S2, [110]online
   supplemental file 2). All patients were included following the
   inclusion criteria. Inclusion criteria: (1) age ≥18 years old; (2)
   patients with MPLC with imaging confirmation; (3) underwent neoadjuvant
   chemoimmunotherapy and subsequent surgery in this center; (4)
   histopathology, clinical, imaging and treatment data were available;
   (5) written informed consent. The pathological response of all resected
   lesions was evaluated by two experienced pulmonary pathologists in
   accordance with International Association for the Study of Lung Cancer
   (IASLC) neoadjuvant pathology recommendations. No viable tumor cell was
   required for pCR. Tumors with a viable percentage exceeding 10% were
   classified as partial responders.

Whole-exome sequencing

   DNA was extracted using the TIANamp DNA extraction kit, and quality
   assessed using agarose electrophoresis, NanoDrop spectrophotometry and
   Qubit V.3.0 fluorometry. Library preparation was performed and
   hybridization capture for Illumina V.2. The process included DNA
   fragmentation, end repair, adaptor ligation, amplification and
   purification steps, with final library quality and concentration
   assessed prior to sequencing on a high-throughput platform using PE150.

RNA sequencing and analysis

   Total RNA was extracted using TRIzol, with RNA Integrity Number (RIN)
   >7.0 required for library construction. Libraries were quantified with
   the KAPA kit, validated by reverse transcription quantitative
   polymerase chain reaction (RT-qPCR), and sequenced on an Illumina
   NovaSeq. RNA sequencing data, aligned to the hg38 genome using HISAT2,
   underwent quality checks with FastQC. Microenvironment Cell Populations
   counter (MCP-counter)[111]^23 analyzed cellular components and immune
   infiltration from the expression matrix. DESeq identified
   Differentially expressed genes (DEGs) with ≥2-fold change and p≤0.05,
   applying false discovery rate (FDR) for correction. Pathway enrichment
   analysis followed. TLS scores were calculated via single-sample Gene
   Set Enrichment Analysis (ssGSEA) using markers references,[112]^24