Background

Invasive lung adenocarcinomas (LUADs) account for almost 70–90% of all surgically resected lung cancers [1]. The morphologic manifestations of invasive LUADs have been well characterized microscopically and are mainly differentiated into lepidic, papillary, acinar, micropapillary, and solid growth patterns [1, 2]. The diversity of histologic composition macroscopically reflects the inter-tumor and intra-tumor heterogeneity of LUADs, with most LUADs manifesting as a successive tissue transition between two or more histologic patterns [2]. Solid and acinar are two histologic subtypes of LUAD with high frequency, the solid type was identified as a histologic pattern stronger in aggressiveness, higher in grade, and worse in prognosis than the acinar type [2]. Identification of patients who may benefit from additional treatment after curative surgery for early LUAD has been a focus in the field of adjuvant therapy. An early study suggested that the solid type benefited from adjuvant chemotherapy in terms of disease-free survival (DFS) and specific DFS, while the acinar type did not [3]. Moreover, the latest grading system, introduced with 20% or more of high-grade patterns (including the solid pattern) as the cut-off for histologically high-grade LUADs, consistently demonstrated that those patients with high-grade LUADs could benefit from adjuvant chemotherapy [Immunohistochemistry

Tissues were fixed in 4% paraformaldehyde, embedded in paraffin, cut into sections, and placed on adhesion microscope slides. Sections were subjected to immunohistochemical (IHC) staining according to standard procedures. We performed the IHC by using the MTHFD2 mouse anti-human antibody (Abcam, ab56772). The primary antibody was incubated at 4 °C overnight followed by using the BOND™ Polymer Refine Detection Kit (Leica, DS9800) according to the manufacturer’s instructions. Whole slide scanning was performed using panoramic MIDI under a 40 × objective lens. For each slide, the histologic patterns were firstly identified according to the cellular structure of the tumor, then three to five non-overlap** fields of view for each histologic region were randomly captured at 100 × magnification, and the staining intensity of MTHFD2 was finally semi-quantified using the Image J software (1.53q) by transforming it into mean optical density [32]. The statistical difference in staining intensity of MTHFD2 between solid and lepidic/acinar was determined by the Wilcoxon rank-sum test.

Statistics

The statistical analyses involved in this study were described in the corresponding method section. All statistical analyses and data presentations were performed by the R program (versions 3.6.3 and 4.0.2). All reported P values were two-tailed, and P < 0.05 was considered statistically significant.

Results

Analysis of scRNA-seq data from histologically annotated LUAD samples

The present study was a repurposing of scRNA-seq data from two of our previously published researches [12, 13]. All surgically excised samples came from patients with untreated, primary non-metastatic LUADs. The histologic constituents of each tumor sample were assessed and recorded semi-quantitatively [1, 2]. We attempted to single out tumor samples with high histologic purity for the purpose of dissecting subtype-specific oncological and immunological characteristics. Collectively, four solid-type, four acinar-type LUAD samples, and five adjacent normal lung samples were enrolled in this study. The representative hematoxylin–eosin (HE) stained images clearly visualized the microscopic structure of the acinar and solid patterns (Additional file 7: Fig. S1A–B). There was a "near-pure" tumor with the solid pattern covering more than 70% of the whole tumor in each solid LUAD sample [33]. With regards to the acinar type, the proportion of acinar pattern in each sample was greater than 50%, with the content of solid/micropapillary patterns limited to less than 10%, allowing to minimize the impacts of high-grade histologic components. The clinicopathological information for all enrolled samples was summarized in Additional file 1: Table S1.

The single-cell transcriptomic profiles generated by each sample were then combined for integrated analysis. Following strict quality control procedures, a sparse matrix with 97,875 cells and 25,233 genes was obtained (Methods). Before performing unsupervised graph-based clustering analysis, potential batch effects between samples were assessed and eliminated. Subsequently, all cells were labeled preliminary based on the expression of canonical cell markers (roughly, PTPRC for immune cells, EPCAM for epitheliums, VWF and COL1A2 for stromal cells; Additional file 7: Fig. S1C-D). Among these cells, 30,208 (30.86%) originated from solid samples, 25,250 (25.80%) originated from acinar samples, and 42,417 (43.34%) originated from adjacent lung tissues.

Tumor cells from solid LUADs create a more anoxic and acidic TME

We then committed to comparing the transcriptional characteristics of tumor cells derived from solid or acinar samples. By inferring large-scale copy number variations from transcriptome information, extensive chromosomal aberrations were observed in tumor-derived epitheliums relative to stromal cells (Additional file 7: Fig S1E). Comparing solid and acinar samples using gene set variation analyses (GSVA) [34] revealed that hallmarks associated with aggressiveness and metabolic activity, such as G2M checkpoint, angiogenesis, epithelial-mesenchymal transition (EMT), MYC targets V1 and PI3K/AKT/mTOR signaling, were up-regulated in tumor cells from solid samples (Fig. 1A, Additional file 2: Table S2), which was consistent with a more aggressive histopathological phenotype of solid LUADs. Notably, immune response-related hallmarks (such as TNFα signaling via NF-κB, IL2-STAT5 signaling, and IL6-JAK-STAT3 signaling) were also significantly enriched in solid samples. These findings emphasized the invasiveness of tumor cells from solid LUADs as well as their adept immune evasion capabilities.

Fig. 1
figure 1

Tumor cells from solid LUADs create a more anoxic and acidic tumor microenvironment. A. Differentially enriched hallmarks (top) and KEGG pathways (bottom) between tumor cells from solid and acinar samples revealed by GSVA. B. Violin plots showing enrichment scores of tumor proliferating rate and hypoxia signatures by histologic subtypes in the TCGA LUAD cohort. Global differences were measured by the Kruskal–Wallis test. C. Violin plots of upregulated genes in solid LUAD tumor cells. D. Boxplots showing mRNA expression of HIF1A, LDHA, UBE2S and UBE2C by histologic subtypes in the TCGA LUAD cohort. Box centerlines, median; box limits, the 25th and 75th percentiles; box whiskers, 1.5 × the interquartile range. Comparisons were performed using two-sided Wilcoxon rank-sum test (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, n s not significant). E–H. Violin plots showing enrichment scores of glycolysis (E), lactate transmembrane transporter activity (F), checkpoint molecules (G) and ubiquitin mediated proteolysis (H) signatures by histologic subtypes in the TCGA LUAD cohort. Global differences were measured by the Kruskal–Wallis test

Full size image

It has been well established that hypoxia and acidification characterized the tumor microenvironment [35]. When comparing tumor cells from solid and acinar samples, hypoxia and glycolysis hallmarks were found to be more prominent in the former (Additional file 2: Table S2). Consistently, by applying single sample enrichment analysis (ssGSEA) in bulk RNA-seq data from the TCGA LUAD cohort, we found the enrichment scores of the tumor proliferating rate [36] and hypoxia [25] signatures were increased stepwise with histologic progression (Fig. 1B; Additional file 2: Table S2). Moreover, the expression levels of hypoxia-inducible factor-1 alpha (HIF1A) and lactate dehydrogenase A (LDHA), were observably upregulated in tumor cells from solid LUADs (Fig. 1C, Additional file 2: Table S2). As the key mediator of hypoxic response, HIF1A was intimately linked to multiple aspects of antitumor immunobiological processes [73]. A concomitant concern is whether the progression of histologic patterns is accompanied by the transformation in metabolic profiles of LUADs. Here we did find a gradient of metabolic alterations and relatively specific metabolic preferences between histologic subtypes, these metabolic properties coincided with the malignant potential of the histologic subtypes and might have a direct or indirect impact on intra-tumoral immune function. Energetically, tumor cells from solid LUADs upregulated glycolytic activity, confronting immune cells, which also relied on glycolysis for effector functions, with a scarcer energy source [10]. Notably, the reasons for the spatial distribution and functional differences of immune cells in the solid histologic region remain elusive. Here we propose the following two potential explanations. Firstly, the potential contribution of differences in the spatial distribution of oxygen and nutrients in the tumor regions of solid LUADs; and secondly, the obstruction by ECM components to the migration and movement of immune cells. In the case of the former, we introduce here the tumor model proposed by Lloyd et al. whereby tumor cores tend to maximize their population density and exhibit static, less proliferative phenotypes, while tumor margins are characterized by aggressive proliferative phenotypes [85]. Intriguingly, this model fits highly with the spatial characteristics of the solid pattern of LUAD identified by Tavernari et al. [10]. The harsh metabolic microenvironment created by vicious competition for limited resources in the tumor core may be detrimental to the survival and functional execution of immune cells. Indeed, Lambrechts et al. also suggested that the degree of hypoxia increases progressively from the tumor margin toward the core, whereas most immune cells are inclined to accumulate at the normoxic tumor margin [69]. In the case of the latter, CAFs and their remodeling of the ECM are key factors in structuring the immune infiltration barrier [86]. Based on comparative analysis of transcriptional profiles of the identified fibroblasts, we find that the fibrillar collagen transcriptional level is significantly higher in solid LUAD-derived fibroblasts [28]. And bulk transcriptome-based analysis further confirmed the elevated fibrillar collagen transcription and extracellular matrix remodeling activities in solid LUADs. In addition, it was noteworthy that solid LUADs are often accompanied by substantial intracellular and extracellular mucus production and secretion [1]. This implies that therapeutic regimens targeting CAFs or local ECM potentially promote immune infiltration into the tumor core of solid LUADs, thereby increasing the inter-contact between immune and tumor compartments.

Conclusions

Collectively, we herein proposed some potential entry points to disrupt the immune exclusion and immunosuppressive phenotype and to potentiate immunotherapeutic efficacy for solid LUADs, yet the realization of these notions requires further investigation and validation at different experimental techniques scales, such as microdissection and spatial omics techniques, as well as tumor models. Furthermore, considering the prospect of possible future applications of histologic subtype-directed LUAD treatment, the development of methods to determine the histologic composition or the presence of certain key components in the tumor prior to treatment is crucial.