Metastasis of human tumors to lymph nodes (LN) is a universally negative prognostic factor. LN stromal cells (SC) play a crucial role in enabling T-cell responses, and because tumor metastases modulate their structure and function, this interaction may suppress immune responses to tumor antigens. The SC subpopulations that respond to infiltration of malignant cells into human LNs have not been defined. Here, we identify distinctive subpopulations of CD90+ SCs present in melanoma-infiltrated LNs and compare them with their counterparts in normal LNs. The first population (CD90+ podoplanin+ CD105+ CD146+ CD271+ VCAM-1+ ICAM-1+ α-SMA+) corresponds to fibroblastic reticular cells that express various T-cell modulating cytokines, chemokines, and adhesion molecules. The second (CD90+ CD34+ CD105+ CD271+) represents a novel population of CD34+ SCs embedded in collagenous structures, such as the capsule and trabeculae, that predominantly produce extracellular matrix. We also demonstrated that these two SC subpopulations are distinct from two subsets of human LN pericytes, CD90+ CD146+ CD36+ NG2 pericytes in the walls of high endothelial venules and other small vessels, and CD90+ CD146+ NG2+ CD36 pericytes in the walls of larger vessels. Distinguishing between these CD90+ SC subpopulations in human LNs allows for further study of their respective impact on T-cell responses to tumor antigens and clinical outcomes.

Cancer cells can enter lymphatic vessels and travel to lymph nodes (LN) where they can form LN metastasis. In normal LNs, different populations of stromal cells (SC) of mesenchymal origin, including fibroblastic reticular cells (FRC), follicular dendritic cells (FDC), and marginal reticular cells (MRC), construct an elaborate LN architecture, supporting various LN functions (1–6). For example, FRCs produce factors that support T-cell survival and form a conduit system delivering small antigens and signaling molecules to immune cells within the LN (7, 8). However, following tumor infiltration into LNs, tumor-subverted SCs contribute to tumor progression by mechanisms like the production of tumor growth factors, stimulation of angiogenesis, and suppression antitumor immune responses (9–11).

Although nine or more distinct nonendothelial SC subsets have been identified in murine LNs (12, 13), it remains unclear how many SC subpopulations are present in human LNs and how they respond to tumor infiltration, in part, due to the technical challenges of isolating SCs ex vivo and the scarcity of human LNs for such studies (14, 15). Earlier work in human LNs revealed the complex heterogeneity within lymphatic endothelial cells and identified the unique characteristics of MRCs, highlighting key differences between murine and human LN SCs (16, 17). Here, we further dissected the diversity within human LN SCs, and identified the presence of additional SC subpopulations (CD34+ SCs and subsets of pericytes) that were previously uncharacterized. We demonstrated the presence of distinct subsets of tumor-associated CD90+ SCs including FRCs and CD34+ SCs in melanoma-infiltrated LNs (MILN), which differ from their counterparts in normal LNs. Finally, we showed that FRCs and CD34+ SCs in MILNs displayed transcriptionally distinctive profiles, although both of them can express fibroblast activation protein (FAP), a widely used marker to identify FRCs in murine LNs and cancer-associated fibroblasts (CAF) in human cancers.

Collectively, our findings demonstrated that the composition of human LN stroma is more complex than previously appreciated and that melanoma infiltration induced changes in LN SCs. Our data also revealed heterogeneity within the SCs in melanoma metastases and provide a molecular map to precisely identify distinctive SC subpopulations.

LN tissue and cell samples

Normal LNs with mild reactive changes (18, 19) were excised from the axillary, inguinal, cervical, mesenteric, and mediastinum regions from 9 donors undergoing surgery. MILNs were obtained from 11 patients with metastatic melanoma undergoing elective surgery, and clinical details of the MILNs studied are available in Supplementary Table S1. Written informed consent from patients or next of kin was obtained in accordance with the Declaration of Helsinki, under protocols approved by the Austin Health Human Research Ethics Committee (Heidelberg, Melbourne, Victoria, Australia) and the Northern Regional Ethics Committee (New Zealand).

To generate a single-cell suspension, fresh LN samples were cut and digested at 37°C for 20 to 60 minutes in RPMI1640 (Gibco-BRL) containing Liberase DH (0.2 mg/mL; Roche) and DNaseI (100 U/mL; Sigma-Aldrich). The enzymes were replenished three times, and the reaction was terminated by adding EDTA to 3.3 mmol/L final concentration. The remaining tissue was mechanically dissociated with gentleMACS Dissociator (Miltenyi Biotec). The enzyme and mechanically generated preparations were combined, passed through a 75-μm cell strainer, and cryopreserved.

Antibodies and reagents

The primary antibodies used for immunofluorescence microscopy are detailed in Supplementary Table S2. Primary antibodies were detected using isotype-specific goat anti-mouse, goat anti-rat, or goat anti-rabbit secondary conjugated to Alexa Fluor 488, 555, or 647 (Invitrogen). Antibodies used for flow cytometry are detailed in Supplementary Table S3.

Multicolor immunofluorescence microscopy

LN tissue samples were embedded in TissueTek OCT Compound (Sakura Finetek), snap-frozen in liquid nitrogen, and sectioned 5-μm thick using a cryostat. Frozen LN sections were fixed with ice-cold acetone for 5 minutes and blocked with 0.25% casein ± 10% human serum for 10 minutes at room temperature. Tissue sections were then probed with primary antibodies for 1 hour at room temperature, and with secondary antibodies and DAPI for 30 minutes at room temperature. Slides were mounted using Prolong Gold (Invitrogen; catalog no. P10144) and visualized with a Leica DMRE (Leica) or Eclipse Ni-U Microscope (Nikon) equipped with the epi-fluorescent filters: UV, 450 to 490, 530 to 560, and 590 to 650 nm. Images were generated using Cytosketch (CytoCode) and quantified using multi-wavelength cell-scoring analysis in MetaMorph V.7.8.10 (Molecular Devices).

Flow cytometry

Frozen LN cell samples were thawed, washed, and rested at 37°C in RPMI1640 containing 10% FBS and 50 U/mL Benzonase Nuclease (Merck) for at least 1 hour prior to staining. Thereafter, cells were incubated with antibodies on ice for 30 minutes. DAPI (1:5,000) was added prior to analysis. Samples were run through a BD FACSAria II and data were analyzed using FlowJo V10.4 (Tree Star). Data are available in the Flow Repository (FR-FCM-Z2L6).

FACS of LN SCs

For FACS isolation of normal LN CD34+ SCs, CD45 SCs were first enriched using the EasySep Human CD45 Depletion Kit (Stemcell Technologies; catalog no. 17898). After staining with antibodies, CD34+ CD90+ CD31 CD45 SCs were sorted into a chamber slide containing DMEM with 10% FBS, using a BD FACSAria II equipped with BD FACSDiva software. Cell doublets were excluded using FCS-A/FSC-H and SSC-A/SSC-H, and DAPI+ dead cells were excluded following the gating method to select the CD45 nonhematopoietic fraction from LN single-cell suspension (Supplementary Fig. S1M). For isolating CD34+ SCs and FRCs from MILNs, CD90+ SCs were first enriched using FITC-conjugated anti-CD90 followed by anti-FITC Microbeads (Miltenyi Biotec). After antibody staining, DAPI live FRCs and CD34+ SCs were sorted into RNAse-free Eppendorf tube containing Lysis Buffer (Qiagen; catalog no. 74004) supplemented with 1% beta-Mercaptoethanol (Sigma-Aldrich). RNA was prepared using the RNeasy Micro Kit (Qiagen; catalog no. 74004), and the quantity and quality of resulting RNA were measured using the Agilent RNA 6000 Pico Assay (Agilent Technologies).

In vitro functional assays

Adipogenic differentiation of isolated CD34+ SCs was performed as described previously (20). Briefly, after culturing CD34+ SCs for 4 days, the medium was replaced with either fresh media consisting of DMEM/F12 and 10% FBS for undifferentiated samples, or adipogenic media consisting of DMEM/F12, 10% FBS, 250 μmol/L IBMX (Sigma), 0.5 μmol/L Dexamethasone (Sigma), 5 μmol/L Insulin (Sigma), and 100 μmol/L Indomethacin (Sigma). Media were replenished every 2 to 3 days for 14 days. Subsequently, cells were fixed with acetone for 5 minutes, blocked with 0.25% casein for 10 minutes, and incubated with anti-FABP4 (Supplementary Table S2) for 1 hour at room temperature. Cells were then probed with anti-rabbit IgG Alexa 488, phalloidin Alexa 555 (Life Technologies), and DAPI (1:2,000) for 30 minutes at room temperature. The ability of FRCs to produce IL7 in the culture supernatant was assessed using the Human IL-7 Quantikine HS ELISA Kit (R&D Systems; catalog no. HS750).

Microarray assay and analysis

Microarray analysis of FRCs and CD34+ SCs isolated from MILNs was performed in collaboration with University of Auckland Genomics Centre, using the Clariom S Human Pico Assay (Applied Biosystems). Transcriptome Analysis Console (TAC; version 4.0, Affymetrix) was used to assess quality control parameters, and perform data normalization and statistical analysis. Data were normalized using the transformation-robust multi-chip analysis (SST-RMA) algorithm in TAC. For pairwise comparative analysis, genes that showed a fold change (FC) greater than 2 (FC > 2) and a Padj (FDR) less than 0.05 were considered significant. Gene ontology (GO) enrichment analysis was performed using the Database for Annotation, Visualization, and Integrated Discovery (DAVID, version 6.8; http://david.abcc.ncifcrf.gov/) and the modified Fisher exact test (21, 22). Promiscuous results, such as probes linked to unmapped genes (n = 4) and those with more than one gene name (n = 13), were excluded. The resulting gene list was uploaded on DAVID using the “official gene symbol” identifier. The default whole-genome background was used to perform the initial enrichment analysis. After the analysis, the annotations with a modified Fisher exact P < 0.01 were selected. Data are available in the Gene Expression Omnibus (GSE150965).

Statistical analysis

Statistical analyses were performed using GraphPad Prism (v7). Correlations between the density of T cells and CD90+ SCs or collagen were assessed by a linear regression analysis. A nominal significance threshold of 0.05 was used for P values. Unpaired t test (two-tailed) was used to compare FAP, CD26, or MFAP5 expressed by the CD34+ SCs or FRCs. P values less than 0.05 were considered significant.

Molecular characteristics of FRCs in normal LNs

FRCs represent the main SC type within the T-cell zones of LN, and are found within the podoplanin (PDPN)+ fraction of CD45 CD31 cells, which also include MRCs and FDCs (16). We first examined the phenotype of FRCs in normal human LNs using immunofluorescent microscopy. LN sections from donors were stained with anti-CD3 and anti-CD21 to locate T- and B-cell zones, respectively (Fig. 1A). Human FRCs that express CD90 (16) were predominantly distributed within the T-cell zones and constructed an intricate reticular network spanning this region (Fig. 1B and C). FRCs in the T-cell zones also expressed PDPN, CD146 (16), vascular cell adhesion protein 1 (VCAM-1), intercellular adhesion molecule 1 (ICAM-1), and alpha smooth muscle actin (α-SMA), and wrapped around the outer layer of blood vessels (Fig. 1BE; Supplementary Fig. S1). In contrast to the dense FRC network in the T-cell zones, B-cell follicles contained a sparse network of FRCs and collagen fibers (Fig. 1B; Supplementary Fig. S1I), consistent with the differences described in murine LNs (7).

Figure 1.

Molecular profile of FRCs in normal human LNs. A–E, Normal LN tissue sections probed with antibodies against CD90, PDPN, CD3, CD21, CD146, and collagen were examined by fluorescence microscopy to assess the phenotype of human LN FRCs. Data are representative of multiple LN sections from at least three independent experiments using different LN donors. Scale bars, 200 μm, except in C (50 μm). B, B-cell follicles; CA, capsule; T zone, T-cell zones; TR, trabeculae. F, Human LN single-cell suspensions were analyzed by polychromatic flow cytometry with a panel of antibodies, including anti-CD45, CD31, PDPN, CD90, CD146, CD105, CD271, CD36, HLA-DR, FAP, CD26, CD73, and CD34. Cell doublets were excluded using FCS-A/FSC-H and SSC-A/SSC-H, and DAPI+ dead cells, were excluded (Supplementary Fig. S1M). From the CD45 nonhematopoietic fraction, FRCs were identified as PDPN+CD31 cells, and their phenotype is shown using contour plots. Quadrant gates were set using fluorescence minus one controls. Data shown are representative of 3 different LN donors. BEC, blood endothelial cell; LEC, lymphatic endothelial cell.

Figure 1.

Molecular profile of FRCs in normal human LNs. A–E, Normal LN tissue sections probed with antibodies against CD90, PDPN, CD3, CD21, CD146, and collagen were examined by fluorescence microscopy to assess the phenotype of human LN FRCs. Data are representative of multiple LN sections from at least three independent experiments using different LN donors. Scale bars, 200 μm, except in C (50 μm). B, B-cell follicles; CA, capsule; T zone, T-cell zones; TR, trabeculae. F, Human LN single-cell suspensions were analyzed by polychromatic flow cytometry with a panel of antibodies, including anti-CD45, CD31, PDPN, CD90, CD146, CD105, CD271, CD36, HLA-DR, FAP, CD26, CD73, and CD34. Cell doublets were excluded using FCS-A/FSC-H and SSC-A/SSC-H, and DAPI+ dead cells, were excluded (Supplementary Fig. S1M). From the CD45 nonhematopoietic fraction, FRCs were identified as PDPN+CD31 cells, and their phenotype is shown using contour plots. Quadrant gates were set using fluorescence minus one controls. Data shown are representative of 3 different LN donors. BEC, blood endothelial cell; LEC, lymphatic endothelial cell.

Close modal

Flow cytometry confirmed that the majority of PDPN+CD31 cells in the LN represented FRCs expressing CD90, CD146, CD105, and CD271 (Fig. 1F).Unlike murine FRCs that express FAP (23), no FAP expression was detected on the majority of human FRCs by flow cytometry (Fig. 1F), which was consistent with the immunofluorescence microscopy results (Supplementary Fig. S1J). Human FRCs were also negative for the expression of CD26 (Fig. 1F), the molecule closely related to FAP (24). In vitro, cultured LN SCs with the FRC phenotype were capable of producing IL7 (Supplementary Fig. S1K and S1L), a factor required for T-cell survival (8), suggesting these cells were likely to play a role in maintaining homeostasis of T cells.

Despite several molecules, such as PDPN and VCAM-1, being expressed by both FRCs in the T-cell zones and FDCs at the center of the follicles, FDCs were negative for CD90 and α-SMA (Fig. 1B; Supplementary Fig. S1F). The majority of SCs located in the LN capsule and trabeculae were distinguishable from the FRCs as they largely lacked the expression of PDPN, CD146, VCAM-1, and ICAM-1 (Fig. 1E; Supplementary Fig. S1A and S1H), although they expressed CD90 and α-SMA (Fig. 1B; Supplementary Fig. S1G). A summary of the phenotype of human LN FRCs relative to other SC subpopulations is given in Supplementary Table S4.

Identification of human LN pericytes

Whereas murine FRCs are present around blood vessels, it remains unknown whether FRCs in human LNs truly represent pericytes, especially given the lack of consensus on the phenotypic definition of pericytes in humans (25). In human LNs, we noted a layer of CD146hi cells surrounding the CD31+ blood vascular endothelium (Fig. 2A) that was distinct from the outer layer of FRCs. This high expression of CD146 was in agreement with the phenotypic trait of pericytes previously identified in multiple human tissues (26). These cells were easily distinguished from the outer FRC layer as they expressed much higher CD146 and lacked PDPN expression (Fig. 2AC). Because of relatively weak CD146 expression on FRCs, it was sometimes difficult to clearly visualize CD146 expression by FRCs using microscopy in perivascular areas, as this often caused oversaturation of the CD146 signal on CD146hi pericytes (Fig. 2A and C; Supplementary Fig. S2A). However, flow cytometry clearly demonstrated CD146 on the surface of FRCs (Figs. 1F and 2B), and confirmed that CD146hiPDPN pericytes represented a distinct population from CD146+PDPN+ FRCs (Fig. 2B).

Figure 2.

Identification of pericytes in human LNs. A, Normal LN sections probed with anti-CD146, CD31, and PDPN reveal the presence of CD146hi pericytes surrounded by an outer layer of FRCs. Data are representative of four independent experiments using different LN donors. Scale bar, 50 μm. B, Flow cytometry analysis of LN pericytes using a panel of antibodies, including anti-CD45, CD31, CD90, CD34, PDPN, CD146, HLA-DR, CD36, CD105, CD271, FAP, CD26, and CD73. After excluding CD31+ endothelial cells from the CD45 fraction of LN cells (Supplementary Fig. S1M), CD146hi PDPN pericytes were identified within the CD90+CD34 SC gate. Quadrant gates were set according to fluorescence minus one controls. Data shown are representative of three independent experiments using different LN donors. C–J, LN sections stained with indicated antibodies show CD146hiCD36+ and CD146hiNG2+ pericytes. PNAd+ vessels indicate HEVs. Images in I and J were acquired from serial sections. Data are representative of at least three independent experiments using different LN donors. Scale bars, 50 μm (C–E) and 100 μm (F–J). CA, capsule; SCS, subcapsular sinus.

Figure 2.

Identification of pericytes in human LNs. A, Normal LN sections probed with anti-CD146, CD31, and PDPN reveal the presence of CD146hi pericytes surrounded by an outer layer of FRCs. Data are representative of four independent experiments using different LN donors. Scale bar, 50 μm. B, Flow cytometry analysis of LN pericytes using a panel of antibodies, including anti-CD45, CD31, CD90, CD34, PDPN, CD146, HLA-DR, CD36, CD105, CD271, FAP, CD26, and CD73. After excluding CD31+ endothelial cells from the CD45 fraction of LN cells (Supplementary Fig. S1M), CD146hi PDPN pericytes were identified within the CD90+CD34 SC gate. Quadrant gates were set according to fluorescence minus one controls. Data shown are representative of three independent experiments using different LN donors. C–J, LN sections stained with indicated antibodies show CD146hiCD36+ and CD146hiNG2+ pericytes. PNAd+ vessels indicate HEVs. Images in I and J were acquired from serial sections. Data are representative of at least three independent experiments using different LN donors. Scale bars, 50 μm (C–E) and 100 μm (F–J). CA, capsule; SCS, subcapsular sinus.

Close modal

LN pericytes showed variable expression of CD105 and CD271, and mostly lacked expression of HLA-DR, FAP, CD26, and CD34 (Fig. 2B). CNN1 (Calponin-1), a marker often used to identify pericytes in murine LNs (2), was only expressed by a small subset of CD146hi pericytes around occasional vessels (Supplementary Fig. S2B). However, a large proportion of the CD146hiPDPN pericytes expressed the scavenger receptor CD36 (Fig. 2B and C), which is absent on FRCs (Fig. 1F). LN blood vessels that were surrounded by CD146+CD36+ pericytes included high endothelial venules (HEV) (Fig. 2CE). In contrast, many vessels penetrating the LN capsule and hilum, and a few vessels within the inner parenchyma, were encapsulated by CD146hi pericytes with high expression of neural/glial antigen 2 (NG2; Fig. 2F and G), a marker of pericytes enclosing arterioles and capillaries, but not venules (27). The pericytes surrounding PNAd+ HEVs were mostly negative for NG2 (Fig. 2H). This observation strongly suggested that NG2 and CD36 mark two distinct pericyte subpopulations surrounding different types of blood vessels. Further costains confirmed that expression of CD36 and NG2 did not generally overlap (Fig. 2I and J; Supplementary Fig. S2C and S2D). Thus, human LNs contained two distinct pericyte subpopulations, CD146hiCD36+ pericytes that wrap around the majority of LN blood vasculature including HEVs, and CD146hiNG2+ pericytes that were mostly found in the capsule and hilum. These pericytes, together with FRCs, likely play important functions in controlling the permeability of the LN blood vasculature.

Characterization of CD34+ SCs in normal human LNs

While analyzing LN SCs using flow cytometry, it was noted that the CD45CD31 fraction of nonhematopoietic and nonendothelial cells contained a SC population that expressed CD34, CD90, CD271, and CD105 (Fig. 3A). These cells were distinct from FRCs in that they express high CD34, but are negative or low for PDPN (Fig. 3A). These CD34+ SCs largely lack expression of HLA-DR, FAP, and CD26 (Fig. 3A). This surface phenotype was consistent with that of mesenchymal progenitor cells previously identified in various human tissues (20, 28). CD34+ SCs isolated from human LN adhered to plastic during culture, proliferated extensively, and displayed a fibroblast-like morphology (Supplementary Fig. S3A). Several changes were observed in the surface phenotype during culture. While consistently expressing CD90, these cells gradually lost CD34 expression and gained low expression of PDPN (Supplementary Fig. S3B). Some of the CD34+ SCs were capable of differentiating into adipocytes in vitro, as shown by their expression of fatty acid binding protein 4 (FABP4; Supplementary Fig. S3C). Cultured CD34+ SCs also highly expressed the receptor activator of NF-κB ligand (RANKL; Supplementary Fig. S3D), a molecule expressed by osteoblasts and their progenitor cells (29) and bone marrow–derived mesenchymal progenitor cells (30).

Figure 3.

Characterization of CD34+ SCs in normal human LNs. A, Polychromatic flow cytometry of LN single-cell suspensions demonstrated the presence of CD34+SCs (red) and their surface phenotype. After excluding CD45+ cells and CD31+ endothelial cells, CD34+CD90+ SCs were identified within the CD31 gate. Quadrant gates were set using fluorescence minus one controls. Data shown are representative of at least three independent experiments using LNs from different donors. B–H, Human LN sections were stained for indicated markers to examine the location and phenotype of CD34+ SCs. Data are representative of at least three independent experiments using different LN samples. Scale bars, 200 μm (B and C) and 100 μm (D–H). CA, capsule; SCS, subcapsular sinus; TR, trabeculae.

Figure 3.

Characterization of CD34+ SCs in normal human LNs. A, Polychromatic flow cytometry of LN single-cell suspensions demonstrated the presence of CD34+SCs (red) and their surface phenotype. After excluding CD45+ cells and CD31+ endothelial cells, CD34+CD90+ SCs were identified within the CD31 gate. Quadrant gates were set using fluorescence minus one controls. Data shown are representative of at least three independent experiments using LNs from different donors. B–H, Human LN sections were stained for indicated markers to examine the location and phenotype of CD34+ SCs. Data are representative of at least three independent experiments using different LN samples. Scale bars, 200 μm (B and C) and 100 μm (D–H). CA, capsule; SCS, subcapsular sinus; TR, trabeculae.

Close modal

We next assessed the distribution of CD34+ LN SCs. As noted in an earlier study (31), fibrous structures such as the capsule, trabeculae, and hilum showed bright CD34 expression (Fig. 3BH), indicating these structures represent the main niche for CD34+ SC. The majority of CD34+ SCs in these structures expressed CD90, CD105, and CD271, but they largely lacked expression of CD31, PDPN, and VCAM-1 (Fig. 3BH), consistent with the phenotype established by flow cytometry (Fig. 3A). Close inspection also revealed that some of the CD34+ SCs were located in the perivascular area surrounding large blood vessels adjacent the capsule and hilum (Fig. 3C and F; Supplementary Fig. S3E and S3F), where they often enclosed a layer of NG2+ pericytes (Supplementary Fig. S3E and S3F). However, such perivascular CD34+ SCs were absent in the inner LN parenchyma where CD34 expression was restricted to blood endothelial cells (BEC; Fig. 3B and C).

Taken together, our results indicated that human LNs contain a population of CD34+ SCs that displayed a molecular profile similar to that of mesenchymal progenitor cells identified in other human tissues, and demonstrated that CD34+ SCs mainly reside in the LN capsule and trabeculae.

Disruption of the stromal architecture in MILNs

When tumor cells spread to LNs, various components of the LN stroma are subverted to promote tumor progression and suppress LN immune responses (32). Several important changes were noted in the architecture of MILNs. Tumor infiltration disrupted the organization of T- and B-cell zones in 10 of 11 MILNs with macrometastasis examined (Fig. 4; Supplementary Fig. S4). Among them, five MILN samples still contained several B- and T-cell zones (Fig. 4A), whereas others showed complete disruption of the entire parenchyma with no clear T- or B-cell zones, but contained small clusters of T and B cells (Supplementary Fig. S4A and S4B). LN fibrosis, as shown by expansion of the extracellular matrix (ECM) and CD90+ SCs (33, 34), was observed in eight of 11 MILN samples with macrometastasis (Fig. 4BE; Supplementary Fig. S4C and S4D). This phenomenon was not only limited to the capsule and trabeculae, but was also found within the inner LN regions spanned by the thicker reticular fiber network (Supplementary Fig. S4D). Tumor clusters were often encapsulated by heavily packed layers of ECM and CD90+ SCs, whereas T cells were largely excluded from the inner tumor cluster but found within the surrounding stroma (Fig. 4BE; Supplementary Fig. S4C). Intratumoral regions mostly devoid of T cells contained a relatively sparse reticular network of ECM and CD90+ SCs (Fig. 4BD). In all MILN samples with no clear T-cell zones, sparsely distributed T cells were mostly colocalized with CD90+ SCs and ECM (Supplementary Fig. S4B). Quantification analysis of the MILN images stained for T cells, ECM, and CD90+SCs confirmed that the density of T cells correlated with that of CD90+ SCs and ECM (Fig. 4F and G).These observations revealed that the overall stromal architecture of MILNs was substantially different from normal LNs.

Figure 4.

Alteration of stromal architecture in MILNs. A–E, MILN sections were probed with PRAME or MART1 antibodies to locate melanoma cells. T and B cells were detected with anti-CD3 and anti-CD20, respectively; ECM were detected with anti-collagen or fibronectin; and cancer-associated SCs were detected using anti-CD90. Images shown in A demonstrate representative features of five independent experiments using different MILNs that retained intact T- and B-cell areas. Images in BE are representative of three independent experiments using different MILNs. Scale bars, 500 μm, except in E (100 μm). F and G, Linear regression analysis of the images (n = 56) stained for CD3, CD90, and collagen. Scatter plots demonstrate correlations between the density of T cells and the density of CD90+ SCs or ECM represented by collagen. A P value less than 0.05 was considered significant.

Figure 4.

Alteration of stromal architecture in MILNs. A–E, MILN sections were probed with PRAME or MART1 antibodies to locate melanoma cells. T and B cells were detected with anti-CD3 and anti-CD20, respectively; ECM were detected with anti-collagen or fibronectin; and cancer-associated SCs were detected using anti-CD90. Images shown in A demonstrate representative features of five independent experiments using different MILNs that retained intact T- and B-cell areas. Images in BE are representative of three independent experiments using different MILNs. Scale bars, 500 μm, except in E (100 μm). F and G, Linear regression analysis of the images (n = 56) stained for CD3, CD90, and collagen. Scatter plots demonstrate correlations between the density of T cells and the density of CD90+ SCs or ECM represented by collagen. A P value less than 0.05 was considered significant.

Close modal

Distinctive subpopulations of cancer-associated SCs in MILNs

We next investigated the changes in LN SCs following melanoma infiltration and compared how SCs in MILNs differed from normal LN SCs. To this end, we examined the SC fraction from MILNs by flow cytometry using the knowledge gained from the analysis of normal LNs. From the nonhematopoietic and nonendothelial fraction, tumor cells were excluded using their negative expression of CD90, because this marker was consistently absent on tumor cells but highly expressed on LN SCs (Fig. 4C and D; Supplementary Fig. S4D). Despite the substantial changes observed in the architecture of MILNs, we demonstrated the presence of three distinct cancer-associated SC populations within the CD90+ fraction, similar to those identified in normal LNs; these included CD34+ SCs, FRCs (CD34PDPN+), and pericytes (CD34PDPNCD146+; Fig. 5A). The frequency of these SC populations across the MILN samples examined by flow cytometry is indicated in Supplementary Table S5.

Figure 5.

Presence of three distinct SC populations in MILNs. A, Single-cell suspensions from MILN were analyzed by polychromatic flow cytometry with a panel of antibodies including anti-CD45, CD31, CD90, CD34, PDPN, CD146, CD105, CD271, HLA-DR, CD36, FAP, and CD26. After excluding CD31+ endothelial cells from the CD45 fraction, tumor cells were excluded by their lack of expression of CD90. A CD146 versus PDPN plot of CD90+ SCs shows the presence of three different SC populations including CD34+ SCs (red), pericytes (purple), and FRCs (blue). Quadrant gates were set based on fluorescence minus one controls. Data shown are representative of eight independent experiments using different MILN donors. Some levels of variability were observed for the expression levels of FAP, CD26, and CD271, as shown in B and Supplementary Fig. S5. B, Bar graphs show the proportion of CD34+ SCs positive for FAP (left) or CD26 (middle), and the proportion of FRC-like cells positive for FAP (right). Mean values with error bars representing SD are shown. ***, P < 0.001; n.s., no significance (unpaired t test, two-tailed). Data are combined from 3 normal LN and 8 MILN donors. C–E, MILN sections were probed with indicated antibodies to assess the distribution and phenotype of CD34+ SCs. Images are representative of five independent experiments using different MILN donors. Scale bars, 200 μm (C), 500 μm (D), and 100 μm (E). F, MILN sections were probed with anti–ICAM-1, VCAM-1, and CD90 to assess the phenotype of FRC–like cells. Data are representative of three independent experiments using different MILN samples, and scale bar, 200 μm.

Figure 5.

Presence of three distinct SC populations in MILNs. A, Single-cell suspensions from MILN were analyzed by polychromatic flow cytometry with a panel of antibodies including anti-CD45, CD31, CD90, CD34, PDPN, CD146, CD105, CD271, HLA-DR, CD36, FAP, and CD26. After excluding CD31+ endothelial cells from the CD45 fraction, tumor cells were excluded by their lack of expression of CD90. A CD146 versus PDPN plot of CD90+ SCs shows the presence of three different SC populations including CD34+ SCs (red), pericytes (purple), and FRCs (blue). Quadrant gates were set based on fluorescence minus one controls. Data shown are representative of eight independent experiments using different MILN donors. Some levels of variability were observed for the expression levels of FAP, CD26, and CD271, as shown in B and Supplementary Fig. S5. B, Bar graphs show the proportion of CD34+ SCs positive for FAP (left) or CD26 (middle), and the proportion of FRC-like cells positive for FAP (right). Mean values with error bars representing SD are shown. ***, P < 0.001; n.s., no significance (unpaired t test, two-tailed). Data are combined from 3 normal LN and 8 MILN donors. C–E, MILN sections were probed with indicated antibodies to assess the distribution and phenotype of CD34+ SCs. Images are representative of five independent experiments using different MILN donors. Scale bars, 200 μm (C), 500 μm (D), and 100 μm (E). F, MILN sections were probed with anti–ICAM-1, VCAM-1, and CD90 to assess the phenotype of FRC–like cells. Data are representative of three independent experiments using different MILN samples, and scale bar, 200 μm.

Close modal

Unlike their counterpart in normal LNs, CD34+ SCs in MILNs consistently expressed PDPN in all the samples examined (Fig. 5A and C). A significantly higher proportion of the CD34+ SCs in MILNs expressed CD26 and FAP compared with the CD34+ SC in normal LNs (Figs. 3A and 5A and B; Supplementary Fig. S5A and S5B). Immunofluorescence microscopy further demonstrated that CD34+ SCs in MILNs were located in the capsule, as well as in areas containing dense ECM (Fig. 5C and D; Supplementary Fig. S5A and S5B).

Whereas the phenotype of FRCs in MILNs remained largely consistent with that of normal FRCs, the proportion of these cells positive for FAP markedly increased in MILNs (Figs. 1F, and 5A, B, and E). FRCs in MILNs often lacked or showed low expression of ICAM-1 and VCAM-1 (Fig. 5F), unlike normal FRCs that highly expressed these markers (Supplementary Fig. S1C and S1D).

In contrast to the CD34+ SCs and FRCs in MILNs, CD146+ pericytes mostly lacked expression of FAP and PDPN (Fig. 5A). As in normal LNs, CD146+pericytes in MILNs could be further divided into CD36+ and CD36 subsets (Fig. 5A; Supplementary Fig. S5C).

Taken together, our data demonstrated the changes in LN SCs following tumor infiltration and the presence of distinct cancer-associated SC populations in MILNs. In particular, there was elevated expression of FAP on CD34+ SCs and FRCs in MILNs when compared with their normal counterparts, strongly suggesting their activation (35).

Transcriptome profile of CD34+ SCs and FRCs in MILNs

Finally, we performed a transcriptomic analysis of CD34+ SCs and FRCs isolated from MILNs to explore the potential differences between these subtypes. FACS-isolated CD34+ SCs and FRCs were directly subjected to microarray, and we identified 411 genes that were differentially expressed (FC > 2; FDR < 0.05) between the CD34+ SCs and FRCs from MILNs. Of these, 254 genes showed significantly higher expression in CD34+ SCs, and 157 genes were significantly higher in FRCs, illustrating the distinct expression profiles of the two cancer-associated SC populations (Fig. 6A and B). CD34, a key marker used for distinguishing these two populations, was expressed more highly in CD34+ SCs (FC = 285.09; FDR = 0.0001), validating the FACS results (Fig. 6B). After applying a higher threshold (FDR < 0.01), 97 genes were found to be significantly upregulated by CD34+ SCs, and 36 genes were upregulated by FRCs. The top 25 genes in each group are shown in Supplementary Tables S6 and S7. Protein expression of an ECM component, microfibril associated protein 5 (MFAP5) was higher in CD34+ SCs than in FRCs in all MILNs examined (Supplementary Fig. S6A and S6C), consistent with the microarray results. CD34+ SCs in normal LNs also expressed higher MFAP5 than FRCs (Supplementary Fig. S6B and S6C), indicating an intrinsic ability of these cells to efficiently produce this ECM protein.

Figure 6.

Differential gene expression between CD34+ SCs and FRCs isolated from MILNs. A, Heatmap representation of hierarchical clustering of differentially expressed genes between CD34+ SCs and FRCs isolated from 7 different MILN donors. The numeric scale indicates the log2 signal intensity per gene. The subcluster contains arrays of CD34+ SCs (left) and arrays of FRCs (right). B, Volcano plot of differentially expressed genes between CD34+ SCs and FRCs isolated from 7 different MILN donors. Colored dots represent the genes that passed the minimum threshold FC > 2 and <−2 (vertical lines), with FDR < 0.05. Green and red dots represent the genes upregulated in CD34+ SCs and FRCs, respectively. The lower horizontal line indicates FDR < 0.05, and the upper horizontal line indicates FDR < 0.01.

Figure 6.

Differential gene expression between CD34+ SCs and FRCs isolated from MILNs. A, Heatmap representation of hierarchical clustering of differentially expressed genes between CD34+ SCs and FRCs isolated from 7 different MILN donors. The numeric scale indicates the log2 signal intensity per gene. The subcluster contains arrays of CD34+ SCs (left) and arrays of FRCs (right). B, Volcano plot of differentially expressed genes between CD34+ SCs and FRCs isolated from 7 different MILN donors. Colored dots represent the genes that passed the minimum threshold FC > 2 and <−2 (vertical lines), with FDR < 0.05. Green and red dots represent the genes upregulated in CD34+ SCs and FRCs, respectively. The lower horizontal line indicates FDR < 0.05, and the upper horizontal line indicates FDR < 0.01.

Close modal

To gain insight into the biological implications of the differentially expressed genes between the two cancer-associated SC types, we performed GO functional enrichment analysis using DAVID. The genes upregulated in CD34+ SCs (FC > 2; FDR < 0.05) were significantly enriched (modified Fisher exact P < 0.01) in 34 GO terms for biological processes. Notably, many of these genes were associated with ECM organization, ECM disassembly, and cell–matrix adhesion (Table 1). A keyword search of published literature identified other genes that were either components of the ECM, or have potential roles in the modulation of the ECM. GO enrichment analysis also revealed genes that were related to positive regulation of osteoblast differentiation and positive regulation of adipocyte differentiation (Table 1). To investigate this further, we examined the list of the genes upregulated in CD34+ SCs and found additional genes that associate with adipogenic (FABP3), osteogenic (OGN and EBF2) and chondrogenic (PRG4) lineages (Supplementary Table S6).

Table 1.

List of differentially expressed genes for CD34+ SCs and FRCs isolated from MILNs.

Biological functionSubtype enrichedPPadjGeneLog2 fold differenceFDR
ECM organization CD34+ SCs 0.00034 0.17 VIT 9.69 0.0012 
    MFAP5 8.68 0.0012 
    NDNF 4.39 0.0433 
    FBN1 4.22 0.0048 
    ELN 2.81 0.0002 
    FBLN1 2.7 0.0082 
    COMP 2.17 0.0142 
    ADAMTSL4 1.86 0.0002 
    LAMA2 1.81 0.0004 
    ITGB3 4.19 0.0092 
    ECM2 3.18 0.0003 
ECM disassembly CD34+ SCs 0.0037 0.27 ADAMTS5 5.67 0.0024 
    FBN1 4.22 0.0048 
    GSN 3.37 0.0125 
    ELN 2.81 0.0002 
    MMP2 2.43 0.0082 
    SH3PXD2B 1.66 0.0167 
Cell–matrix adhesion CD34+ SCs 0.0075 0.37 CD34 8.15 0.0001 
    ITGB3 4.19 0.0092 
    ECM2 3.18 0.0003 
    TIAM1 3.17 0.0303 
    PXN 2.84 0.0046 
    SNED1 2.23 0.0456 
Positive regulation of fat cell differentiation CD34+ SCs 0.0037 0.26 LMO3 7.12 0.0002 
    SFRP2 4.13 0.0067 
    MEDAG 4.09 0.0276 
    SH3PXD2B 1.66 0.0167 
    BMP2 1.37 0.0276 
Positive regulation of osteoblast differentiation CD34+ SCs 0.0088 0.39 IGF1 4.71 0.0012 
    SFRP2 4.13 0.0067 
    IL6 3.47 0.0049 
    JAG1 1.44 0.0276 
    BMP2 1.37 0.0276 
Cellular response to tumor necrosis factor FRCs 0.00025 0.044 VCAM1 4.92 0.0118 
    POSTN 4.81 0.0168 
    IL18BP 2.78 0.0142 
    HDAC1 1.47 0.0276 
    CCL21 7.29 0.0039 
    CCL2 1.68 0.015 
    CCL19 4.9 0.0019 
Inflammatory response FRCs 0.0034 0.17 CCL21 7.29 0.0039 
    CXCL9 5.75 0.0015 
    CCL19 4.9 0.0019 
    CXCL13 4.44 0.0045 
    F11R 4.33 0.0077 
    IL15 3.32 0.033 
    RARRES2 2.84 0.0189 
    C4A 2.79 0.0385 
    C4B 2.79 0.0385 
    CCL2 1.68 0.015 
Immune response FRCs 1.02E-09 1.10E-06 CCL21 7.29 0.0039 
    CXCL9 5.75 0.0015 
    CCL19 4.9 0.0019 
    CXCL13 4.44 0.0045 
    CD74 4.42 0.0075 
    IRF8 4.23 0.0033 
    IL15 3.32 0.033 
    CTSS 3.29 0.0048 
    TNFSF13B 3.05 0.008 
    ADAMDEC1 2.99 0.0456 
    HLA-DRA 2.84 0.0101 
    C7 2.75 0.0344 
    HLA-DRB1 2.58 0.0179 
    HLA-F 2.34 0.0082 
    HLA-B 2.09 0.0125 
    IL7R 1.98 0.0413 
    HLA-C 1.82 0.0082 
    CCL2 1.68 0.015 
    IL27RA 1.5 0.0421 
    HLA-DQA1 1.14 0.0446 
Cell chemotaxis FRCs 0.00017 0.036 CCL21 7.29 0.0039 
    CXCL9 5.75 0.0015 
    VCAM1 4.92 0.0118 
    CCL19 4.9 0.0019 
    CCL2 1.68 0.015 
    LEF1 1.65 0.0458 
Antigen processing and presentation FRCs 4.85E-06 0.0017 CD74 4.42 0.0075 
    CTSS 3.29 0.0048 
    HLA-DRA 2.84 0.0101 
    HLA-DRB1 2.58 0.0179 
    HLA-B 2.09 0.0125 
    HLA-C 1.82 0.0082 
    HLA-DQA1 1.14 0.0446 
Biological functionSubtype enrichedPPadjGeneLog2 fold differenceFDR
ECM organization CD34+ SCs 0.00034 0.17 VIT 9.69 0.0012 
    MFAP5 8.68 0.0012 
    NDNF 4.39 0.0433 
    FBN1 4.22 0.0048 
    ELN 2.81 0.0002 
    FBLN1 2.7 0.0082 
    COMP 2.17 0.0142 
    ADAMTSL4 1.86 0.0002 
    LAMA2 1.81 0.0004 
    ITGB3 4.19 0.0092 
    ECM2 3.18 0.0003 
ECM disassembly CD34+ SCs 0.0037 0.27 ADAMTS5 5.67 0.0024 
    FBN1 4.22 0.0048 
    GSN 3.37 0.0125 
    ELN 2.81 0.0002 
    MMP2 2.43 0.0082 
    SH3PXD2B 1.66 0.0167 
Cell–matrix adhesion CD34+ SCs 0.0075 0.37 CD34 8.15 0.0001 
    ITGB3 4.19 0.0092 
    ECM2 3.18 0.0003 
    TIAM1 3.17 0.0303 
    PXN 2.84 0.0046 
    SNED1 2.23 0.0456 
Positive regulation of fat cell differentiation CD34+ SCs 0.0037 0.26 LMO3 7.12 0.0002 
    SFRP2 4.13 0.0067 
    MEDAG 4.09 0.0276 
    SH3PXD2B 1.66 0.0167 
    BMP2 1.37 0.0276 
Positive regulation of osteoblast differentiation CD34+ SCs 0.0088 0.39 IGF1 4.71 0.0012 
    SFRP2 4.13 0.0067 
    IL6 3.47 0.0049 
    JAG1 1.44 0.0276 
    BMP2 1.37 0.0276 
Cellular response to tumor necrosis factor FRCs 0.00025 0.044 VCAM1 4.92 0.0118 
    POSTN 4.81 0.0168 
    IL18BP 2.78 0.0142 
    HDAC1 1.47 0.0276 
    CCL21 7.29 0.0039 
    CCL2 1.68 0.015 
    CCL19 4.9 0.0019 
Inflammatory response FRCs 0.0034 0.17 CCL21 7.29 0.0039 
    CXCL9 5.75 0.0015 
    CCL19 4.9 0.0019 
    CXCL13 4.44 0.0045 
    F11R 4.33 0.0077 
    IL15 3.32 0.033 
    RARRES2 2.84 0.0189 
    C4A 2.79 0.0385 
    C4B 2.79 0.0385 
    CCL2 1.68 0.015 
Immune response FRCs 1.02E-09 1.10E-06 CCL21 7.29 0.0039 
    CXCL9 5.75 0.0015 
    CCL19 4.9 0.0019 
    CXCL13 4.44 0.0045 
    CD74 4.42 0.0075 
    IRF8 4.23 0.0033 
    IL15 3.32 0.033 
    CTSS 3.29 0.0048 
    TNFSF13B 3.05 0.008 
    ADAMDEC1 2.99 0.0456 
    HLA-DRA 2.84 0.0101 
    C7 2.75 0.0344 
    HLA-DRB1 2.58 0.0179 
    HLA-F 2.34 0.0082 
    HLA-B 2.09 0.0125 
    IL7R 1.98 0.0413 
    HLA-C 1.82 0.0082 
    CCL2 1.68 0.015 
    IL27RA 1.5 0.0421 
    HLA-DQA1 1.14 0.0446 
Cell chemotaxis FRCs 0.00017 0.036 CCL21 7.29 0.0039 
    CXCL9 5.75 0.0015 
    VCAM1 4.92 0.0118 
    CCL19 4.9 0.0019 
    CCL2 1.68 0.015 
    LEF1 1.65 0.0458 
Antigen processing and presentation FRCs 4.85E-06 0.0017 CD74 4.42 0.0075 
    CTSS 3.29 0.0048 
    HLA-DRA 2.84 0.0101 
    HLA-DRB1 2.58 0.0179 
    HLA-B 2.09 0.0125 
    HLA-C 1.82 0.0082 
    HLA-DQA1 1.14 0.0446 

Note: The association between the list of differentially expressed genes and biological functions was analyzed using DAVID (version 6.8). After initial enrichment analysis, the annotations with a modified Fisher exact P < 0.01 were selected. The Padj, as calculated using Benjamini–Hochberg procedure, are shown for each annotation.

GO enrichment analysis of the genes upregulated in FRCs (FC > 2; FDR < 0.05) showed that these genes were significantly enriched (modified Fisher exact P < 0.01) for 30 biological processes including the cellular response to TNF, the inflammatory response, and immune response (Table 1). In particular, several genes were related to antigen processing and presentation and cell chemotaxis (Table 1). A keyword search of published literature identified additional genes (CD82 and FDCSP) related to immune function (Supplementary Table S7). These results demonstrated distinctive transcriptional profiles for CD34+ SCs and FRCs in MILNs, and indicated that CD34+ SCs were mainly involved in ECM organization, whereas FRCs played roles in LN immune responses. It is unclear whether this finding indicates intrinsic differences between FRCs and CD34+ SCs, or distinct responses induced by tumor infiltration, as we were unable to isolate sufficient numbers of SCs from normal LNs to perform transcriptional analysis.

Cancer cell invasion of LNs modifies SC populations and causes pathologic phenomena such as fibrosis (32, 33). These changes are thought to facilitate tumor growth and metastasis (9), yet they have not been characterized in detail, in part, due to lack of knowledge of the different SC subpopulations within human LNs. Here, we defined distinctive SC subpopulations present in MILNs that were likely to play distinctive roles in their interactions with both cancer cells and immune cells. Facile identification of these SC subpopulations now opens the way to studying their respective influence on prognosis of cancers following LN metastasis and their impact on responses to therapy such as checkpoint blockade.

Our data in MILNs were predicated on detailed studies of the SCs within normal LNs. Using CD90 as a marker of the nonendothelial and nonhematopoietic SCs, we were able to distinguish three major groups of CD90+ SCs, FRCs, pericytes, and CD34+ SCs. This study extended our previous findings on SC subpopulations present in normal human LNs (16, 17) and characterized new subsets in LNs including CD34+ SCs and two subpopulations of pericytes.

Among the CD90+ SCs in normal human LNs, FRCs construct a reticular network understood to support the migration of T cells (36). Our data showed that human FRCs expressed many markers in common with FDCs including PDPN, but were distinguished from FDCs by their expression of CD90, CD146, and α-SMA. A sparse network of FRCs and ECM was also present in B-cell follicles, consistent with follicular conduits, which provides a transport route for small antigens and signaling molecules to traffic to FDCs, as described in mice (37, 38). As well as forming a reticular network, these FRCs often formed the outer layer of blood vessels, outside the layer formed by CD146-bright pericytes adjacent to BECs.

Pericytes that envelope BECs bear many characteristics of other fibroblastic SCs, so defining markers that discriminate them from other subsets are needed, especially for flow cytometry where spatial relationships are lost. Although CD146 proved a highly expressed marker of pericytes (26), CD146 was also weakly expressed by FRCs, so the lack of PDPN expression was necessary to accurately distinguish LN pericytes that were CD90+CD146+PDPNCD34. We also discovered that LN pericytes were composed of two subpopulations distinguishable by their expression of CD36 or NG2, respectively. CD36+ pericytes surrounded the majority of blood vessels including HEVs, whereas NG2+ pericytes enveloped the vessels in the capsule and hilum, but were mostly absent from HEVs. These findings will enable new studies of distinctive pericyte populations in other human tissues.

A third major population of CD90+ SCs in human LNs expressed CD34, and these cells had many of the phenotypic characteristics of mesenchymal progenitor cells identified in other human tissues (20, 28, 39, 40). These cells were distinct from FRCs, since they had high CD34 expression but mostly lacked expression of PDPN, ICAM-1, and VCAM-1. CD34+ SCs predominantly resided in fibrous structures such as the capsule and trabeculae, consistent with an earlier observation of CD34+ cells at those sites (31). Some CD34+ SCs were also present in the perivascular region surrounding large blood vessels penetrating the capsule and hilum, consistent with the main location of CD34+ mesenchymal progenitor cells in other tissues (20, 28, 39, 40). Therefore, blood vessels in these areas appear to be surrounded by at least two distinct layers of SCs, first NG2+ pericytes, then CD34+ SCs.

In vitro, CD34+ SCs displayed features common to mesenchymal progenitor cells, such as adherence to plastic, a fibroblast-like morphology, and a high proliferative capacity (41). They also differentiated into adipocytes and expressed RANKL, a molecule reportedly expressed by mesenchymal progenitor cells (29, 30). We did not carry out additional histochemical assays of “tri-lineage” differentiation in vitro (41), because such traditional assays are not reliable measures of multi-lineage potential, especially when performed on cells adapted to cell culture (42). However, we note that PDPNCD31 SCs from human LNs have previously been positive in all three assays, as were PDPN+CD31 FRCs (15). Whereas other methods will be needed to examine the true differentiation potential of the CD34+ SCs, our results suggested that CD34+ SCs were likely to provide a replenishing source of cells to remodel SC networks in LNs during immunologic challenges. In our experiments, CD34+ SCs lost CD34 expression during in vitro culture but gained PDPN expression, ultimately resembling FRCs. In mice, CD34+ SCs differentiate into multiple SC types that can resemble FRCs or pericytes (43, 44).

Using the knowledge gained from our analysis of normal LNs, we investigated how the stromal architecture was altered in MILNs. Eight of 11 MILNs with macrometastasis showed features of fibrosis with expansion of CD90+ SCs, consistent with an earlier study which reported a presence of prominent ECM structures in LNs with macrometastasis (9). We then demonstrated that the three distinct CD90+ SC populations we found in normal LNs were also present in MILNs. However, compared with their normal counterparts, both FRCs and CD34+ SCs in MILNs, but not pericytes, had greatly elevated expression of FAP (35). FAP is expressed by CAFs that have been implicated in suppressing T-cell function (45), and altering how T cells interact with such FAP+ CAFs is now an immunotherapeutic strategy for cancer (46). Our finding that FRCs and CD34+ SCs in MILNs express FAP suggests a possibility that both populations were activated following melanoma infiltration.

We also observed that CD34+ SCs in MILNs, but not FRCs, upregulated expression of CD26. CD26 expression is associated with elevated potential to produce ECM and induce fibrosis (47), suggesting that CD34+ SCs in MILNs may have more fibrogenic potential than FRCs, consistent with their localization to fibrous regions. The low expression of ICAM-1 and VCAM-1 by FRCs in MILNs was likely to modulate their interactions with lymphocytes (1).

Transcriptomic analysis supported the phenotypic diversity of SCs in MILNs, revealing distinctive differences in gene expression profiles of CD34+ SCs and FRCs present in MILNs. CD34+ SCs in MILNs expressed genes associated with ECM organization, confirming these cells have more ECM modulatory potential, consistent with a recent study of normal murine LNs (12). CD34+ SCs in MILNs also showed higher expression of genes related to adipogenic, osteogenic, and chondrogenic lineages, supporting the concept that they may be precursor cells capable of generating different SC subtypes. In contrast, the genes highly expressed by FRCs in MILNs include those involved in antigen presentation and secretion of chemokines that are crucial for LN function (e.g., CCL19, CCL21, and CXCL13; ref. 2). FRCs within MILNs therefore have the transcriptional capacity to modulate lymphocyte trafficking through altering chemokinesis or chemotaxis. In pancreatic tumors, there are two distinct CAF populations within the TME, myofibroblastic and inflammatory CAFs (48). Despite the similarities between CD34+ SCs and myofibroblastic CAFs, and FRCs and inflammatory CAFs, their relationships need further investigation.

Our data also demonstrated that T cells were frequently excluded from the tumor clusters and remained associated with ECM and cancer-associated CD90+ SCs. The close association of T cells with ECM was noted in previous studies of MILNs (9, 33), where the density of infiltrating leukocytes correlated with the amount of intratumoral collagen. The significance of SCs in determining the position of T cells in the TME has been suggested in studies of lung (49), pancreatic (46), and breast cancers (50). It remains unclear whether densely packed ECM layers and SCs represents a physical barrier for T-cell movement (51), or whether tumor cells produce T-cell repulsive chemokines (46). It is also possible that certain SC types express chemokines and adhesion molecules that promote T-cell attachment (33, 52), or additional factors present in hypoxic and nutrient-deficient TMEs contribute to the exclusion of T cells from tumor nests (51). Further studies will determine how different SC populations control the migratory behavior of T cells within MILNs.

In summary, our results revealed distinct subpopulations of CD90+ SCs in human LNs and provided specific molecular markers to identify them. Of these, two subpopulations of CD90+ SCs appeared to become activated following melanoma infiltration and may have potential to modulate T-cell activity in different ways. The distinctive set of markers we have identified has altered the study of both cancer-associated SC subsets. Although, we have only studied MILNs to date, the molecular markers we described open the way to studying their impact in other human tumors. They will also enable functional studies of their interactions with immune cells, which may suggest new therapeutic strategies to enhance the immune response against cancer cells that infiltrate LNs.

J. Eom reports grants from Auckland Medical Research Foundation and Maurice Wilkins Centre during the conduct of the study. S.M. Park reports grants from Healthy Research Council of New Zealand and Maurice Wilkins Centre during the conduct of the study. J.A. Mathy reports grants from Maurice Wilkins Centre and Health Research Council of New Zealand during the conduct of the study. A.E.S. Brooks reports grants from Maurice Wilkins Centre during the conduct of the study. P.R. Dunbar reports grants from New Zealand Tertiary Education Commission, New Zealand Health Research Council, and Auckland Medical Research Foundation during the conduct of the study, as well as grants from Gilead Sciences and grants and personal fees from SapVax LLC outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

J. Eom: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft. S.M. Park: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft. V. Feisst: Resources, data curation, investigation, methodology, writing–review and editing. C.-J.J. Chen: Resources, investigation. J.E. Mathy: Resources, methodology. J.D. McIntosh: Methodology, writing–review and editing. C.E. Angel: Methodology, writing–review and editing. A. Bartlett: Resources. R. Martin: Resources. J.A. Mathy: Resources. J.S. Cebon: Resources. M.A. Black: Formal analysis, supervision, methodology, writing–review and editing. A.E.S. Brooks: Conceptualization, data curation, supervision, validation, methodology, writing–review and editing. P.R. Dunbar: Conceptualization, resources, formal analysis, supervision, funding acquisition, methodology, writing–original draft, project administration, writing–review and editing.

For donated clinical material, the authors thank the staff and patients of the Austin Hospital (Melbourne, Victoria, Australia) and hospitals of Waitemata and Manukau district health boards (Auckland, New Zealand). The authors are grateful to Dr. Otto Strauss for help with sample collection. This work was supported by grants from the Maurice Wilkins Centre, Auckland Medical Research Foundation, and Healthy Research Council of New Zealand.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Mueller
SN
,
Germain
RN
. 
Stromal cell contributions to the homeostasis and functionality of the immune system
.
Nat Rev Immunol
2009
;
9
:
618
29
.
2.
Malhotra
D
,
Fletcher
AL
,
Astarita
J
,
Lukacs-Kornek
V
,
Tayalia
P
,
Gonzalez
SF
, et al
Transcriptional profiling of stroma from inflamed and resting lymph nodes defines immunological hallmarks
.
Nat Immunol
2012
;
13
:
499
510
.
3.
Malhotra
D
,
Fletcher
AL
,
Turley
SJ
. 
Stromal and hematopoietic cells in secondary lymphoid organs: partners in immunity
.
Immunol Rev
2013
;
251
:
160
76
.
4.
Roozendaal
R
,
Mebius
RE
. 
Stromal cell-immune cell interactions
.
Annu Rev Immunol
2011
;
29
:
23
43
.
5.
Katakai
T
,
Suto
H
,
Sugai
M
,
Gonda
H
,
Togawa
A
,
Suematsu
S
, et al
Organizer-like reticular stromal cell layer common to adult secondary lymphoid organs
.
J Immunol
2008
;
181
:
6189
200
.
6.
Chang
JE
,
Turley
SJ
. 
Stromal infrastructure of the lymph node and coordination of immunity
.
Trends Immunol
2015
;
36
:
30
9
.
7.
Sixt
M
,
Kanazawa
N
,
Selg
M
,
Samson
T
,
Roos
G
,
Reinhardt
DP
, et al
The conduit system transports soluble antigens from the afferent lymph to resident dendritic cells in the T cell area of the lymph node
.
Immunity
2005
;
22
:
19
29
.
8.
Link
A
,
Vogt
TK
,
Favre
S
,
Britschgi
MR
,
Acha-Orbea
H
,
Hinz
B
, et al
Fibroblastic reticular cells in lymph nodes regulate the homeostasis of naive T cells
.
Nat Immunol
2007
;
8
:
1255
65
.
9.
Soikkeli
J
,
Podlasz
P
,
Yin
M
,
Nummela
P
,
Jahkola
T
,
Virolainen
S
, et al
Metastatic outgrowth encompasses COL-I, FN1, and POSTN up-regulation and assembly to fibrillar networks regulating cell adhesion, migration, and growth
.
Am J Pathol
2010
;
177
:
387
403
.
10.
Pereira
ER
,
Jones
D
,
Jung
K
,
Padera
TP
. 
The lymph node microenvironment and its role in the progression of metastatic cancer
.
Semin Cell Dev Biol
2015
;
38
:
98
105
.
11.
Kalluri
R
. 
The biology and function of fibroblasts in cancer
.
Nat Rev Cancer
2016
;
16
:
582
98
.
12.
Rodda
LB
,
Lu
E
,
Bennett
ML
,
Sokol
CL
,
Wang
X
,
Luther
SA
, et al
Single-cell RNA sequencing of lymph node stromal cells reveals niche-associated heterogeneity
.
Immunity
2018
;
48
:
1014
28
.
13.
Pezoldt
J
,
Pasztoi
M
,
Zou
M
,
Wiechers
C
,
Beckstette
M
,
Thierry
GR
, et al
Neonatally imprinted stromal cell subsets induce tolerogenic dendritic cells in mesenteric lymph nodes
.
Nat Commun
2018
;
9
:
3903
.
14.
Fletcher
AL
,
Malhotra
D
,
Acton
SE
,
Lukacs-Kornek
V
,
Bellemare-Pelletier
A
,
Curry
M
, et al
Reproducible isolation of lymph node stromal cells reveals site-dependent differences in fibroblastic reticular cells
.
Front Immunol
2011
;
2
:
35
.
15.
Severino
P
,
Palomino
DT
,
Alvarenga
H
,
Almeida
CB
,
Pasqualim
DC
,
Cury
A
, et al
Human lymph node-derived fibroblastic and double-negative reticular cells alter their chemokines and cytokines expression profile following inflammatory stimuli
.
Front Immunol
2017
;
8
:
141
.
16.
Park
SM
,
Angel
CE
,
McIntosh
JD
,
Brooks
AE
,
Middleditch
M
,
Chen
CJ
, et al
Sphingosine-1-phosphate lyase is expressed by CD68 cells on the parenchymal side of marginal reticular cells in human lymph nodes
.
Eur J Immunol
2014
;
44
:
2425
36
.
17.
Park
SM
,
Angel
CE
,
McIntosh
JD
,
Mansell
CM
,
Chen
CJ
,
Cebon
J
, et al
Mapping the distinctive populations of lymphatic endothelial cells in different zones of human lymph nodes
.
PLoS One
2014
;
9
:
e94781
.
18.
Luscieti
P
,
Hubschmid
T
,
Cottier
H
,
Hess
MW
,
Sobin
LH
. 
Human lymph node morphology as a function of age and site
.
J Clin Pathol
1980
;
33
:
454
61
.
19.
Elmore
SA
. 
Histopathology of the lymph nodes
.
Toxicol Pathol
2006
;
34
:
425
54
.
20.
Feisst
V
,
Brooks
AE
,
Chen
CJ
,
Dunbar
PR
. 
Characterization of mesenchymal progenitor cell populations directly derived from human dermis
.
Stem Cells Dev
2014
;
23
:
631
42
.
21.
Huang
DW
,
Sherman
BT
,
Lempicki
RA
. 
Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources
.
Nat Protoc
2009
;
4
:
44
57
.
22.
Huang
DW
,
Sherman
BT
,
Lempicki
RA
. 
Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists
.
Nucleic Acids Res
2009
;
37
:
1
13
.
23.
Denton
AE
,
Roberts
EW
,
Linterman
MA
,
Fearon
DT
. 
Fibroblastic reticular cells of the lymph node are required for retention of resting but not activated CD8+ T cells
.
Proc Natl Acad Sci U S A
2014
;
111
:
12139
44
.
24.
Havre
PA
,
Abe
M
,
Urasaki
Y
,
Ohnuma
K
,
Morimoto
C
,
Dang
NH
. 
The role of CD26/dipeptidyl peptidase IV in cancer
.
Front Biosci
2008
;
13
:
1634
45
.
25.
Armulik
A
,
Genove
G
,
Betsholtz
C
. 
Pericytes: developmental, physiological, and pathological perspectives, problems, and promises
.
Dev Cell
2011
;
21
:
193
215
.
26.
Crisan
M
,
Yap
S
,
Casteilla
L
,
Chen
C-W
,
Corselli
M
,
Park
TS
, et al
A perivascular origin for mesenchymal stem cells in multiple human organs
.
Cell Stem Cell
2008
;
3
:
301
13
.
27.
Murfee
WL
,
Skalak
TC
,
Peirce
SM
. 
Differential arterial/venous expression of NG2 proteoglycan in perivascular cells along microvessels: identifying a venule-specific phenotype
.
Microcirculation
2005
;
12
:
151
60
.
28.
Zimmerlin
L
,
Donnenberg
VS
,
Pfeifer
ME
,
Meyer
EM
,
Peault
B
,
Rubin
JP
, et al
Stromal vascular progenitors in adult human adipose tissue
.
Cytometry A
2010
;
77
:
22
30
.
29.
Xiong
J
,
Onal
M
,
Jilka
RL
,
Weinstein
RS
,
Manolagas
SC
,
O'Brien
CA
. 
Matrix-embedded cells control osteoclast formation
.
Nat Med
2011
;
17
:
1235
41
.
30.
Huang
H
,
Kim
HJ
,
Chang
EJ
,
Lee
ZH
,
Hwang
SJ
,
Kim
HM
, et al
IL-17 stimulates the proliferation and differentiation of human mesenchymal stem cells: implications for bone remodeling
.
Cell Death Differ
2009
;
16
:
1332
43
.
31.
Diaz-Flores
L
,
Gutierrez
R
,
Garcia
MP
,
Saez
FJ
,
Diaz-Flores
L
 Jr
,
Valladares
F
, et al
CD34+ stromal cells/fibroblasts/fibrocytes/telocytes as a tissue reserve and a principal source of mesenchymal cells. Location, morphology, function and role in pathology
.
Histol Histopathol
2014
;
29
:
831
70
.
32.
Jones
D
,
Pereira
ER
,
Padera
TP
. 
Growth and immune evasion of lymph node metastasis
.
Front Oncol
2018
;
8
:
36
.
33.
Samaniego
R
,
Estecha
A
,
Relloso
M
,
Longo
N
,
Escat
JL
,
Longo-Imedio
I
, et al
Mesenchymal contribution to recruitment, infiltration, and positioning of leukocytes in human melanoma tissues
.
J Invest Dermatol
2013
;
133
:
2255
64
.
34.
Fletcher
AL
,
Acton
SE
,
Knoblich
K
. 
Lymph node fibroblastic reticular cells in health and disease
.
Nat Rev Immunol
2015
;
15
:
350
61
.
35.
Huber
MA
,
Kraut
N
,
Park
JE
,
Schubert
RD
,
Rettig
WJ
,
Peter
RU
, et al
Fibroblast activation protein: differential expression and serine protease activity in reactive stromal fibroblasts of melanocytic skin tumors
.
J Invest Dermatol
2003
;
120
:
182
8
.
36.
Bajenoff
M
,
Egen
JG
,
Koo
LY
,
Laugier
JP
,
Brau
F
,
Glaichenhaus
N
, et al
Stromal cell networks regulate lymphocyte entry, migration, and territoriality in lymph nodes
.
Immunity
2006
;
25
:
989
1001
.
37.
Bajenoff
M
,
Germain
RN
. 
B-cell follicle development remodels the conduit system and allows soluble antigen delivery to follicular dendritic cells
.
Blood
2009
;
114
:
4989
97
.
38.
Roozendaal
R
,
Mempel
TR
,
Pitcher
LA
,
Gonzalez
SF
,
Verschoor
A
,
Mebius
RE
, et al
Conduits mediate transport of low-molecular-weight antigen to lymph node follicles
.
Immunity
2009
;
30
:
264
76
.
39.
Simmons
PJ
,
Torok-Storb
B
. 
CD34 expression by stromal precursors in normal human adult bone marrow
.
Blood
1991
;
78
:
2848
53
.
40.
Lin
CS
,
Xin
ZC
,
Deng
CH
,
Ning
H
,
Lin
G
,
Lue
TF
. 
Defining adipose tissue-derived stem cells in tissue and in culture
.
Histol Histopathol
2010
;
25
:
807
15
.
41.
Dominici
M
,
Le Blanc
K
,
Mueller
I
,
Slaper-Cortenbach
I
,
Marini
F
,
Krause
D
, et al
Minimal criteria for defining multipotent mesenchymal stromal cells. The International Society for Cellular Therapy position statement
.
Cytotherapy
2006
;
8
:
315
7
.
42.
Robey
P
. 
"Mesenchymal stem cells": fact or fiction, and implications in their therapeutic use
.
F1000Research
2017
;
6
:
524
.
43.
Sitnik
KM
,
Wendland
K
,
Weishaupt
H
,
Uronen-Hansson
H
,
White
AJ
,
Anderson
G
, et al
Context-dependent development of lymphoid stroma from adult CD34(+) adventitial progenitors
.
Cell Rep
2016
;
14
:
2375
88
.
44.
Benezech
C
,
Mader
E
,
Desanti
G
,
Khan
M
,
Nakamura
K
,
White
A
, et al
Lymphotoxin-β receptor signaling through NF-κB2-RelB pathway reprograms adipocyte precursors as lymph node stromal cells
.
Immunity
2012
;
37
:
721
34
.
45.
Yang
X
,
Lin
Y
,
Shi
Y
,
Li
B
,
Liu
W
,
Yin
W
, et al
FAP promotes immunosuppression by cancer-associated fibroblasts in the tumor microenvironment via STAT3-CCL2 signaling
.
Cancer Res
2016
;
76
:
4124
35
.
46.
Feig
C
,
Jones
JO
,
Kraman
M
,
Wells
RJ
,
Deonarine
A
,
Chan
DS
, et al
Targeting CXCL12 from FAP-expressing carcinoma-associated fibroblasts synergizes with anti-PD-L1 immunotherapy in pancreatic cancer
.
Proc Natl Acad Sci U S A
2013
;
110
:
20212
7
.
47.
Mah
W
,
Jiang
G
,
Olver
D
,
Gallant-Behm
C
,
Wiebe
C
,
Hart
DA
, et al
Elevated CD26 expression by skin fibroblasts distinguishes a profibrotic phenotype involved in scar formation compared to gingival fibroblasts
.
Am J Pathol
2017
;
187
:
1717
35
.
48.
Ohlund
D
,
Handly-Santana
A
,
Biffi
G
,
Elyada
E
,
Almeida
AS
,
Ponz-Sarvise
M
, et al
Distinct populations of inflammatory fibroblasts and myofibroblasts in pancreatic cancer
.
J Exp Med
2017
;
214
:
579
96
.
49.
Salmon
H
,
Franciszkiewicz
K
,
Damotte
D
,
Dieu-Nosjean
MC
,
Validire
P
,
Trautmann
A
, et al
Matrix architecture defines the preferential localization and migration of T cells into the stroma of human lung tumors
.
J Clin Invest
2012
;
122
:
899
910
.
50.
Cremasco
V
,
Astarita
JL
,
Grauel
AL
,
Keerthivasan
S
,
MacIsaac
K
,
Woodruff
MC
, et al
FAP delineates heterogeneous and functionally divergent stromal cells in immune-excluded breast tumors
.
Cancer Immunol Res
2018
;
6
:
1472
85
.
51.
Peranzoni
E
,
Rivas-Caicedo
A
,
Bougherara
H
,
Salmon
H
,
Donnadieu
E
. 
Positive and negative influence of the matrix architecture on antitumor immune surveillance
.
Cell Mol Life Sci
2013
;
70
:
4431
48
.
52.
Tirosh
I
,
Izar
B
,
Prakadan
SM
,
Wadsworth
MH
,
Treacy
D
,
Trombetta
JJ
, et al
Dissecting the multicellular ecosystem of metastatic melanoma by single-cell RNA-seq
.
Science
2016
;
352
:
189
96
.