Spatial transcriptomics study of Castleman disease

Abstract

Background

Castleman Disease (CD) is a rare lymphoproliferative disorder of uncertain etiology, primarily characterized by painless lymphadenopathy. To further understand the possible pathogenesis of the disease, we utilized spatial transcriptomics to explore CD.

Methods

Spatial transcriptomics was performed on FFPE samples from lymph nodes of three CD patients with different histologic types and one normal lymph node.

Results

The sample was categorized by analysis into 14 clusters, including B cells, plasma cells, BECs, LECs, CD4 + T cells, CD8 + T cells, macrophages, monocytes, cycling cells, myofibroblasts, mDCs, NKCs, Tfh and NESCs. Our study showed that the lesion cells of CD might be B cells. In addition to this, we found that mesenchymal cells, especially myofibroblasts, play an important role in disease progression and there was a large heterogeneity of cellular communication between cell clusters in different histologic types of CD.

Conclusions

Through this study we have gained a better understanding of the pathogenesis of CD. It provides new ideas for the subsequent exploration of CD and directions for the development of new clinical drugs.

Background

Castleman Disease can be divided into unicentric castleman disease (UCD) and multicenter castleman disease (MCD) according to clinical manifestations. Based on histomorphologic changes, CD can be classified into hyaline vascular variant (HV), plasma cell variant (PC), and mixed variant [1, 2]. Although we have gained a better understanding of the clinical manifestations and histologic changes of CD, the pathogenesis of CD is still unknown. Previous studies have shown that the abnormal changes of interleukin-6 (IL-6) may be related to the pathogenesis of idiopathic MCD (iMCD), and targeted therapy against IL-6 has been approved for iMCD [3,4,5]. However, blocking IL-6 is only effective for a portion of patients [3, 5]. Therefore, it is urgent to clarify the pathogenesis of CD and provide clues for finding new treatment schemes.

With the use of high-throughput sequencing technology, more and more diseases are having breakthroughs [6,7,8,9]. Spatial transcriptomics (ST) sequencing is a new high-throughput sequencing technique [10]. Compared with traditional transcriptome sequencing, which can only obtain gene expression data, spatial transcriptome sequencing can simultaneously obtain the spatial position information of cells and the gene expression data of cells in corresponding positions. It provides an important research method for many fields such as tissue and cell function, microenvironment interaction, pedigree tracking of development process, disease pathology and so on [11, 12]. In the previous study of CD, clinical classification was mainly used [13], while in this study, considering that the diagnosis of CD mainly depends on the basis of pathology, it was planned to explore pathogenesis from the perspective of histological classification. Therefore, we utilized spatial transcriptome technology on the 10X Genomics platform to study one normal lymph node and three lesion lymph nodes which were from CD patients with hyaline vascular, plasma cell, and mixed types, respectively. In this study, we not only characterized the spatial information that was lacking in previous studies of CD, but also further explored the pathogenesis of CD in a profound way.

Methods

Human patient samples

The clinical data of patients diagnosed with CD in the Department of Pathology of our hospital from June 2021 to May 2023 were collected. Since CD is an autoimmune disease, later associated with the onset of some malignancies [14,15,16,17,18], it is necessary to avoid the impact of other related diseases on the CD sequencing results as much as possible. The final decision on the inclusion criteria were: (1) patients with definite diagnosis (diagnosed by two or more pathologists according to the World Health Organization guidelines); (2) patients without other autoimmune diseases and malignant tumors; and (3) typical pathological structure. There were 4 samples. One sample of the most typical formalin fixation and paraffin embedding (FFPE) lymph node was selected from each of the three histologic types, and one sample of normal FFPE lymph node was selected as a control. In each FFPE sample, we manually selected a 6.5 × 6.5 mm area (the most typical lesion structure) based on its microscopic morphology for spatial transcriptomic study. The study was approved by the Ethics Committee and was conducted in accordance with the Declaration of Helsinki.

Spatial transcriptomics

Slide preparation

RNA quality of the tissue is assessed by calculating DV200 of RNA extracted from freshly collected tissue sections. The tissue block with exposed tissue is trimmed or scored to make it compatible in size to the Capture Areas on Visium slides. The tissue block is then sectioned by a microtome to generate appropriately sized sections for Visium slides. Sections are collected in a water bath and are allowed to float on the water surface until they are flat and free from wrinkles and folds. Expanded sections are then placed within the frames of Capture Areas on Visium Spatial slides.

Section deparaffinization, staining and imaging

Slides were incubated uncovered at 60 °C for 2 h. Then, slides were cooled down to room temperature. Dip the slide into xylene in xylene Jar 1 and xylene in xylene Jar 2 in turn, and incubate for 10 min respectively. Next gently immerse slide in the Ethanol. Finally, immerse slide in the water in the tube and indulge for 20 s. For staining, the sections were incubated in hematoxylin for 3 min, bluing buffer for 1 min, and alcoholic eosin solution for 1 min. After each staining step, sections were washed with water. And then, image all Capture Areas individually at the desired magnification using brightfield imaging settings.

Tissue optimization (TO)

Briefly, the Visium Spatial Tissue Optimization workflow includes placing tissue sections on 7 Capture Areas on a Visium Tissue Optimization slide. The sections are fixed, stained, and then permeabilized for different times. mRNA released during permeabilization binds to oligonucleotides on the Capture Areas. Fluorescent cDNA is synthesized on the slide and imaged. The permeabilization time that results in maximum fluorescence signal with the lowest signal diffusion is optimal. If the signal is the same at two time points, the longer permeabilization time is considered optimal.

Libraries construction

Transcriptome probe panel, consisting of a pair of specific probes for each targeted gene, is added to the tissues. Together, probe pairs hybridize to their complementary target RNA. After hybridization, a ligase is added to seal the junction between the probe pairs that have hybridized to RNA, forming a ligation product. The single stranded ligation products are released from the tissue upon RNase treatment and permeabilization, and then captured on the Visium slides. Once ligation products are captured, probes are extended by the addition of UMI, Spatial Barcode and partial Read 1. The cDNA labeled with Spatial Barcode will be used to infer later on the original spatial position of each RNA molecule. The spatially barcoded, ligated probe products are released from the slide and harvested for qPCR to determine Sample Index PCR cycle number. The products then undergo indexing via Sample Index PCR. This, in turn, generates final library molecules that are cleaned up by SPRIselect, assessed on a bioanalyzer, quantified, and then sequenced.

Data processing

We use in-house script to perform basic statistics of raw data and evaluate the data quality and GC content along the sequencing cycles. In addition, we evaluated the distribution of sequencing quality values, the distribution of sequencing error rate, Q20, Q30 content and other indicators. Raw FASTQ files and histology images were processed by sample with the Space Ranger (version spaceranger-1.2.0, 10X Genomics) software with default parameters. The filtered gene-spots matrix and the fiducial-aligned low-resolution image was used for down-streaming data analyses (Seurat).

Seurat analysis

The Seurat package was used to perform gene expression normalization, dimensionality reduction, spot clustering, and differential expression analysis. Briefly, spots were filtered for minimum detected gene count of 100 genes. Normalization across spots was performed with the SCTransform function and 3000 highly variable genes was selected for principal component analysis. For spot clustering, the first 20 PCs were used to build a graph, which was segmented with a resolution of 0.5. Wilcox algorithm was used to perform differential gene expression analysis for each cluster via FindAllMarkers function. Genes with fold change (FC) > 2 and adjust P < 0.05 were defined as significantly differential expressed genes.

Differential expression gene analysis

The edgeR package is used for differential expression gene analysis [19]. By comparing one cluster of the sample with other clusters of the sample, a list of differential genes between this cluster and other clusters is obtained. Similarly, the same cluster of normal tissues and pathological tissues were compared to obtain the corresponding differential gene list. Based on the differential genes, we screened the genes with higher expression in this cluster than other clusters as candidate marker genes, that is logFC > 0.

Enrichment analysis

The clusterProfiler R package was used to calculate enrichment test for candidate gene sets based on hypergeometric distribution [20]. Pathways with corrected P < 0.05 were considered as significantly enriched terms. Three pathway classification systems were used as reference databases for samples, including Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) and Reactome Pathway. GO and Reactome annotations were retrieved from Ensembl BioMart and KEGG pathways was retrieved via KEGG REST API.

Cell type annotation

We used the GSVA R package for scoring [21]. The genes were ranked according to the cumulative density distribution of the genes in the expression profile matrix, and a rank statistic similar to the K-S test was computed for each gene set, and the expression matrix was converted into a gene set enrichment scoring (ES) matrix to obtain the GSVA enrichment scoring for each gene set corresponding to each cluster. The cell type with the highest score for each cluster was used for annotation.

Copy number variation analysis

After reviewing the relevant literature, we selected NK cells as reference cells and analyzed datas using the inferCNV R package according to the official guidelines. Cells of Other clusters were compared with NK cells to identify those with large copy number variations and each cell for the extent of CNV signal was also scored to finally arrive at the most likely lesion cells.

Trajectory analysis

We used the monocle2 R package to analyze cell development trajectories. The genes that determine cell progression were selected using dpFeature program, and then the reduceDimension function was used to perform dimensionality reduction analysis and constructed the initial cell trajectory. Next we used the orderCells function to sort the cells. Finally, B cells were used as root to present the direction of the whole cell development trajectory.

Cell chat analysis

Cell Chat Analysis was performed using CellChat R package [22]. We base our calculations on the official workflow. Data input, processing and initialization were performed first. Then, screening for receptor-ligand interactions was performed with reference to the CellChatDB Human database. The potential ligand-receptor interactions between lesion and non-lesion cells and potential pathways were calculated using computeCommunProb, computeCommunProbPathway and aggregateNet functions with standard parameters.

Results

Basic clinical characteristics of patients with CD

This study included four FFPE samples, one from normal lymph node and three from lymph nodes of patients with different histologic types of CD, respectively. The lesion tissues were all negative for human herpes virus 8 (HHV-8), and the patients were clinically typed as MCD, excluding TAFRO subtype and POEMS-associated MCD. Serum β2-microglobulin (β2-MG) levels were mildly elevated in the patients with hyaline vascular variant-Castleman Disease (HV-CD), whereas they were normal in the remaining two cases. Elevated serum IL-6 and IL-8 levels were seen in patients with HV-CD and plasma cell variant-Castleman Disease (PC-CD). Patients with PC-CD had the most severe clinical symptoms, which mainly included: loss of weight, fatigue, anemia, thrombocytosis, hypoalbuminemia and elevated C-reactive protein (CRP) (Table 1).

Table 1 Basic information

Cell type annotation and difference between samples

In the four FFPE samples, we captured a total of 17,684 spots, of which 4992 spots were of plasma cell type, 4881 spots were of mixed type, 4967 spots were of hyaline vascular type, and 2844 spots were of normal lymph node tissue. The diameter of each spot is 55 μm, which can capture 1–10 cells (depending on the cell size). We performed gene set variation analysis (GSVA) on the samples and selected the cluster with the highest GSVA score to annotate the spot, and finally divided each FFPE sample into 14 clusters (Supplementary Tables 1–4), including B cells, plasma cells, blood vascular endothelial cells (BECs), lymphatic endothelial cells (LECs), CD4 + T cells, CD8 + T cells, macrophages, monocytes, cycling cells, myofibroblasts, myeloid dendritic cells (mDCs), natural killer cells (NKCs), T follicular helper cells (Tfh) and non-endothelial stem cells (NESCs) (Fig. 1A). Most of the various types of cells in the lymph node tissue showed diffuse distribution, only the B cells in the germinal center were more aggregated, so their spatial structure was more obvious. Comparing the spatial position of the B cells cluster with the microscopic structure after HE staining, we found that the cell type annotation was more accurate (Fig. 1B).

Fig 1

Cell type annotation and cell proportions of Castleman's disease tissues. A Cell type annotation results for samples. BEC blood vascular endothelial cells, NESC non-endothelial stem cells, LEC lymphatic endothelial cells. B Comparison of the spatial position of hyaline vascular variant tissues with HE-stained microscopic structures. The spatial location of B cells is shown on the left, and HE staining is shown on the right. C Cell proportions between samples. The first figure on the left shows the proportion of each cluster in the plasma cell type CD lymph node tissue. The proportion of each cluster in the mixed type CD lymph node tissue is shown in the second figure from the left. The second figure on the right reveals the proportion of each cluster in the hyaline vascular type CD lymph node tissue. The distribution of each cluster in normal lymph node tissue is shown in the first figure on the right.

In order to find out more visually whether there were differences in the proportion of cells with different histological types of CD, we performed a comparison between samples and the results showed that the percentage of B cells of HV-CD was the highest, accounting for 21.1%, and the percentage of NESCs was the highest among the other three samples. The proportion of plasma cells in lymph nodes of PC-CD and mixed CD was significantly higher than that of normal lymph node. At the same time, the proportion of myofibroblasts increased significantly in all lesion tissues, while the proportion of macrophages, monocytes and BECs decreased (Fig. 1C).

Identification and trajectory of lesion cells

Analysis of large-scale chromosome copy number variations (CNV) in somatic cells, such as the gain or loss of entire chromosomes or large segments of chromosomes, can assist in distinguishing malignant from non-malignant cell populations in tissues [23]. Spatial transcriptome sequencing, by integrating CNV information, spatial gene expression, and cell type localization information, can identify microclones that are not morphologically obvious and clarify the transition from benign to malignant tissues [24]. We selected NK cells as reference cells, because it is currently believed that such cells are the least likely to be lesion cells [13, 25], and highest score was B cells after CNV analysis of each cluster (Fig. 2A and Supplementary Tables 5–8). Therefore, we concluded that B cells might be lesion cells.

Fig 2

Identification of lesion cells and trajectory of tissue development. A Copy number variations analysis results. The left panel shows the CNV analysis results of the diseased lymph node tissue (hyaline vascular type as an example), and the right panel shows the CNV analysis results of the normal lymph node tissue. B Differences in the number of cells in normal versus diseased lymph node tissue in advanced stages of development. The first row is the number of plasma cells, myofibroblasts, LECs and cycling cells in advanced stages of development in normal lymph node tissue. The following row shows the number of plasma cells, myofibroblasts, LECs and cycling cells in advanced stages of development in diseased lymph node tissue (hyaline vascular type as an example).C Genes highly expressed at nodes during tissue development. The figure on the left shows the genes that are predominantly highly expressed at the nodes in normal lymph node tissue. The middle and right panels show genes that are highly expressed at nodes during the development of lesion tissue.

To further explore the tissues development process, we performed pseudo-time series analysis on each sample. Taking the B cells as the original cells, we observed there were some differences in the development process of clusters, especially in late stages of development. In all lesion tissues, the number of plasma cells, cycling cells, LECs and myofibroblasts was significantly increased in the late developmental phase (Fig. 2B).

There were multiple stages throughout tissue development, each with a different dominant cells, which was largely determined by gene expression. We called the points where the cell state changed as nodes, and combined with the analysis of the highly expressed genes at different nodes, eventually we got an overall view of the development of lymph node tissues. In normal lymph node tissues, the highly expressed genes at each node were mainly immune-related genes, such as IGHG1, IGHA1 and IGKC. In lesion tissues, not only immune-related genes but also genes related to cell migration and adhesion, such as TMSB10, CEMIP and CD209, were highly expressed at the nodes (Fig. 2C). Therefore, we deduced that the lesional B cells secreted substances that promoted both the proliferation of immune cells and stromal cells, suggesting that stromal cells may play a vital role in the development of CD.

Genetic, functional heterogeneity within lesion tissues

In order to further clarify the related genetic, functional and metabolic alterations during disease development, we performed differential expression gene analysis and enrichment analysis between lesion and normal tissues and between B cells cluster and other clusters of the same sample. Among the many differential genes calculated, we selected top10 genes for comparison. Comparing between lesion and normal tissues, we found that the expression of B2M and IGKC were up-regulated in the B cells cluster of all lesion tissues (Figs. 3A and S1). The remaining clusters of lesion tissues highly expressed differential genes similar to the B cells cluster. After GO and KEGG enrichment analyses, we found that all lesion tissues had enhanced functions in cytoplasm, nucleus, RNA binding, transferase activity, mitochondrion, nucleotide binding, ATP binding, and extracellular exosome. Meanwhile, there were metabolic increases of Protein processing in endoplasmic reticulum, Autophagy-animal, Ubiquitin mediated proteolysis, Metabolic pathways, Lysosome, Neurotrophin signaling pathway and Apoptosis, and an increased likelihood of contracting other diseases such as Salmonella, Shigellosis, Hepatitis B, Yersinia and Chronic myeloid leukemia (Fig. 3B).

Fig 3

Results of differential and enrichment analysis of samples. A Genes with up-regulated expression in all lesion tissues. B Results of enrichment analysis. The left panel shows the results of Gene Ontology (GO) enrichment of the lesion tissue, and the right panel shows the results of Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis (plasma cell type as an example). C Results of Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of B cells in the same sample (plasma cell type as an example).

Next was a comparison between the B cells cluster and other clusters in the same sample, where we found Fc gamma R-mediated phagocytosis, B cell receptor signaling pathway, T cell receptor signaling pathway, C-type lectin receptor signaling pathway, and NF-kappa B signaling pathway were increased in all B cells cluster of lesion tissues (Fig. 3C).

Molecular mechanism related to disease development

For the purpose of better understand the molecular mechanisms associated with disease development and guiding the study of new drugs, we performed cell chat analysis. The results suggested that in normal lymph node tissues, the B cells cluster was most strongly associated with other clusters, especially the NESCs cluster (Figs. 4A and S2 A). From the analysis of input and output signaling patterns, we can see that PRL signaling axis plays an important role in the self-regulation of B cells (Fig. 4E). In PC-CD lymph node tissues, the connection between mDCs and LECs, and between LECs and plasma cells was closer (Fig. 4B). IL-10 pathway was over-activated, and plasma cells acted as an important receiver of this pathway, which we hypothesized to have an important relationship with causing plasma cells in CD to over-proliferate (Figs. 4F and S2 B). In addition to this, we also found that BECs act as signal senders of the BMP and VEGF pathways to plasma cells, which may also be associated with plasma cells hyperproliferation (Fig. 4G, H). In mixed CD lymph node tissues, plasma cells communicated most frequently with myofibroblasts (Fig. 4C). Myofibroblasts became the predominant signal receivers regulated by a variety of signaling pathways, mainly including GRN, MSTN and ANNEXIN signaling pathways (Fig. 4K). Meanwhile, the CD70 signaling pathway had a non-negligible role for plasma cells and cycling cells (Fig. 4I). In HV-CD tissues, all clusters were more closely linked to each other (Fig. 4D). Notably, the B cells cluster was inactive, both as a signal receiver and sender. CSF signaling pathway expression was significantly up-regulated in other clusters except the B cells cluster (Fig. 4J) and the main player in this pathway was CSF1-CSF1R (Figure S2 C).

Fig 4

Cell chat analysis results. A Results of cell chat analysis in normal lymph node tissue. B Results of cell chat analysis in plasma cell type CD tissue. C Results of cell chat analysis in mixed type CD tissue. D Results of cell chat analysis in hyaline vascular type CD tissue. E PRL signaling pathway in normal lymph node tissue. F IL-10 pathway in plasma cell type CD tissue. G BMP pathway in plasma cell type CD tissue. H VEGF pathway in plasma cell type CD tissue. I CD70 pathway in mixed type CD tissue. J CSF pathway in hyaline vascular type CD tissue. K Incoming signal patterns in mixed type CD tissue. L Difference in the CXCL signaling pathway in normal versus lesion tissues. The upper left figure demonstrates the CXCL pathway in normal lymph node tissue. The lower left panel shows the CXCL pathway in the diseased lymph node tissue (plasma cell type as an example). The diagram on the right reflects the receptors and ligands that play a major role in the CXCL pathway in diseased tissue (plasma cell type as an example).

Previous studies have shown that CXCL, VEGF, and PDGF pathways may play important roles in CD [25,26,27], and in this study we found that in normal lymph node tissues, the main signal senders of the CXCL pathway were myofibroblasts and the signal receivers were B cells, whereas in lesion tissues the main signal senders and receivers were B cells, and the role was mainly played by CXCL12-CXCR4 (Fig. 4L). The major signal senders and receivers of the VEGF pathway in lesion tissues varied, but all were associated with BECs, and the PDGF pathway was strongly associated with myofibroblasts (Figure S2 D).

Discussion

This study explored different histologic subtypes of CD by spatial transcriptome technology, exposing their spatial heterogeneity and possible pathogenesis. It is different from conventional transcriptome and single-cell transcriptome sequencing techniques because the utilization of spatial transcriptome technology not only ensures high resolution, but also restores the spatial location of various types of cells, which provides an advantage in exploring the alterations in the disease microenvironment and the relationship between various cellular clusters [10,11,12]. After analysis, we classified each lymph node tissue sample into 14 clusters, and by comparing the proportions of cells in the different subtypes, we found that the proportion of myofibroblasts was increased in all lesion tissues, suggesting that this type of cells may play a crucial role in the development of CD.

CNV revealed that B cells might be the lesion cells. Subsequently, in the pseudo-time series analysis, we found that the number of plasma cells, cycling cells, LECs, and myofibroblasts at the late stage of development was significantly increased when B cells were used as the originating cells in the lesion tissues. We hypothesize that the increase in the number of myofibroblasts may be due to the fact that the lesional B cells promote their proliferation through certain pathways, and then, the stromal cells can also promote the proliferation of other cells through certain pathways, such as the PDGF pathway. As for the increase in cycling cells, it may be an indication of excessive cell proliferation, while the increase in the number of plasma cells is more likely to be the result of an excessive immune response produced by the diseased B cells. LECs may occur as an effector outcome, and may also play a contributing role in the development of the disease. In a study back in 2019, Li et al. found platelet-derived growth factor receptor b (PDGFRB) mutations in 17% of UCDs, which were localized to CD45- cells, possibly representing stromal cells, and confirmed by in vitro experiments that the mutation is a gain of function with proliferative and survival advantages [25]. In a recent study exploring iMCD using single-cell sequencing in conjunction with spatial transcriptomics techniques, the results suggested that the IL-6 pathway was dominant in endothelial cells and fibroblasts [28]. Myofibroblasts, as a type of stromal cell, have been shown to be an important component of the tumor microenvironment and played an important role in the development of a variety of tumors [29,30,31,32,33]. Therefore, it is strongly hypothesized that myofibroblasts assume an important role in the pathogenesis of CD, which also provides a new direction for clinical treatment.

Meanwhile, after analyzing the highly expressed genes at each node during tissue development, we found that in normal lymph node tissues, the highly expressed genes at each node were mainly immune-related genes, while in lesion tissues, the highly expressed genes at the nodes were not only immune-related genes, but also genes related to cell migration and adhesion. This reinforces the fact that stromal cells, which may act as effector cells for diseased B cells, play an important role in the development of CD.

By differential expression gene analysis and enrichment analysis, we found that both B2M and IGKC were up-regulated in the lesion tissues, and a query of gene function using the GeneCards database [34] suggested that B2M is a protein-coding gene with a product of β2-MG, and IGKC is an immune-related gene. This indicates that dynamic detection of β2-MG in serum may be instructive for disease diagnosis and prognostic assessment. Elevated β2-MG was proposed as one of the significant risk factors for CD in a study back in 2018 [35]. The high expression of immune genes proves that immune hyperreactivity is a key cause of CD, and treatment of the disease requires suppression of its immune response. In recent years, anti-IL-6 targeted therapy has achieved better efficacy in CD, especially iMCD, suggesting that inflammatory factors in the immune response may be a breakthrough in treatment. At the same time, it should be cautioned that the results of enrichment analysis suggest that CD patients have an increased likelihood of contracting other diseases. This has been confirmed in previous studies, where CD patients have an increased likelihood of developing lymphoma and Kaposi's sarcoma at a later stage, which may be related to their immunocompromised function due to depletion of immune cells in the body [14, 18, 36, 37]. After comparing the B cells cluster with other clusters in the same samples, we found that the B cells cluster of the lesion tissues had increased metabolism of the NF-kappa B signaling pathway, which regulates genes involved in immunity, inflammation, and cell survival, and which plays a key role in regulating the immune response to infections [38]. Abnormal regulation of this pathway has been associated with malignancy, inflammation, pathogenic microbial infections, and abnormalities in immune development, and it has been shown that when tumor cells activate this pathway, tumor cell proliferation is enhanced and insensitive to apoptosis regulation [39]. Previous studies have found that activation of this pathway may be associated with the development of HHV-8-MCD [40]. Therefore, this pathway may become a new target for targeted therapy.

Unlike previous studies, we did not find a close correlation between IL-6 and the disease in this study. This is not surprising; although it has long been shown that IL-6 is closely associated with the development of CD, and anti-IL-6 targeted drugs have become the treatment of choice for iMCD, there were also studies suggesting that the development of iMCD may not be related to IL-6, as there were no significantly elevated levels of IL-6 in these patients [3, 41]. Drugs targeting IL-6 improve symptoms in only one-third to one-half of patients [3, 5]. Of the three patients enrolled in this study, serum IL-6 levels were mildly elevated in patients with HV-CD, markedly elevated in patients with PC-CD, and normal in patients with mixed CD, none of whom were enriched in the IL-6 pathway before treatment. The heterogeneity of the clinical manifestations of CD and, perhaps, the pathogenesis of the different subtypes of CD are also highly variable. This was suggested in a study in 2022, where sequencing of UCD, MCD, and non-CD tissues revealed that the genes and pathways that were highly expressed in different subtypes were not the same [13]. Similar conclusions were obtained in this study by cell chat analysis. Plasma cell type CD may be associated with the IL-10 pathway, and previous studies have shown that IL-10 is not only associated with survival, proliferation, and anti-apoptotic activity in various cancers, such as Burkitt's lymphoma, non-Hodgkin's lymphoma, and small-cell lung cancer, but also plays an important role in the development of autoimmune diseases [42,43,44,45]. Mixed CD may be associated with the CD27-CD70 signaling pathway, which has been shown to influence immune regulation in the tumor microenvironment, leading to immune escape [46]. The CSF pathway is more active in HV-CD, and the main player in this pathway is CSF1-CSF1R, which has been shown to promote M2 polarization of macrophages, and M2 macrophages are thought to have pro-oncogenic effects [47], so targeting and inhibiting this pathway may have a therapeutic effect on HV-MCD. For the pathways of CXCL, VEGF and PDGF, which have been highly focused on in previous studies, the CXCL pathway reflected a significant abnormality in the present study. In normal lymph node tissues, the main signal senders of the CXCL pathway were myofibroblasts, and the signal receivers were B cells, whereas in lesion tissues the main signal senders and receivers were B cells, and the major players were CXCL12-CXCR4. It is known that CXCL12 is a chemokine, which is mainly expressed and secreted by stromal cells, and when combined with CXCR4, it can induce directional migration of target cells, enhance the adhesion between target cells and endothelial cells, and involve in physiological functions such as cell growth, development, differentiation, apoptosis, etc., so it plays an important role in the growth and metastasis, immune response, and leukocyte activation of tumor cells [48]. The signal sender of this pathway has changed from myofibroblasts to B cells, so it can be speculated that it plays a pivotal role in the development of CD.

Of course, the study has some limitations; firstly, the sample size of the study was small, with only 1 lymph node of each histologic type, which requires a subsequent large sample size study. Secondly, the clinical typing was all MCD, and data on UCD were lacking. However, our study is relatively novel, exposing for the first time the spatial location of each cluster in CD lymph node tissues, as well as the genes, cell types, and pathway alterations in different histologic subtypes of CD, which provides new potential targets for clinical treatment.

Conclusions

Our study showed that the lesion cells of CD might be B cells. In addition to this, we found that mesenchymal cells, especially myofibroblasts, play an important role in disease progression and there was a large heterogeneity of cellular communication between cell clusters in different histologic types of CD. It provides new ideas for the subsequent exploration of CD pathogenesis and directions for the development of new clinical drugs.