Enhanced Lymph Vessel Density, Remodeling, and Inflammation Are Reflected by Gene Expression Signatures in Dermal Lymphatic Endothelial Cells in Type 2 Diabetes

  1. Brigitte Hantusch1
  1. 1Clinical Institute of Pathology, Medical University of Vienna, Vienna, Austria
  2. 2Department of Internal Medicine, Division of Rheumatology, Medical University of Vienna, Vienna, Austria
  3. 3Section for Science of Complex Systems, Medical University of Vienna, Vienna, Austria
  4. 4Division of Vascular Surgery, Department of Surgery, Medical University of Vienna, Vienna, Austria
  1. Corresponding author: Brigitte Hantusch, brigitte.hantusch{at}meduniwien.ac.at.


Type 2 diabetes is associated with microvascular damage that causes frequent infections in the skin and chronic ulcers as a result of impaired wound healing. To trace the pathological changes, we performed a comprehensive analysis of lymphatic vessels in the skin of type 2 diabetic versus nondiabetic patients. The dermis revealed enhanced lymphatic vessel density, and transcriptional profiling of ex vivo isolated lymphatic endothelial cells (LECs) identified 160 genes differentially expressed between type 2 diabetic and nondiabetic LECs. Bioinformatic analysis of deregulated genes uncovered sets functionally related to inflammation, lymphatic vessel remodeling, lymphangiogenesis, and lipid and small molecule transport. Furthermore, we traced CD68+ macrophage accumulation and concomitant upregulation of tumor necrosis factor-α (TNF-α) levels in type 2 diabetic skin. TNF-α treatment of LECs and its specific blockade in vitro reproduced differential regulation of a gene set that led to enhanced LEC mobility and macrophage attachment, which was mediated by the LEC-derived chemokine CXCL10. This study identifies lymph vessel gene signatures directly correlated with type 2 diabetes skin manifestations. In addition, we provide evidence for paracrine cross-talk fostering macrophage recruitment to LECs as one pathophysiological process that might contribute to aberrant lymphangiogenesis and persistent inflammation in the skin.

The incidence of type 2 diabetes and obesity is rapidly increasing worldwide (1). Currently, generalized insulin insensitivity is considered the central pathogenic event (2) that is frequently linked to a systemic metabolic syndrome, a state of chronic low-level inflammation involving macrophage activation in adipose tissue (3). Recent insights also indicate genetic factors in the development of the disease (4). However, chronic hyperglycemia induces extensive macro- and microvascular alterations (5) that lead to systemic organ damage. Large vessels react to the chronically increased glucose and glucose-driven metabolites with enhanced arteriosclerosis. Diabetic microangiopathy gradually evokes retinopathy, leading to subsequent blindness, and nephropathy, the most frequent reason of renal insufficiency. In the skin, microvasculopathy causes prolonged inflammation, impaired healing of wounds, and ulcers (6).

Type 2 diabetes–induced microvascular lesions are characterized by aberrant matrix component deposition, resulting in narrowing of the vascular lumen that causes ischemia. Concurrently, the affected blood vessel endothelium shows imbalances of vasoconstrictors and -dilators, secretion of pro- and anti-inflammatory cytokines, and increased prothrombotic activity (7), which leads to leakiness and sustained effusion of leukocytes and plasma components into the tissue. In contrast to blood vessels, nothing is known so far about the involvement of the lymphatic vasculature in human type 2 diabetes, although dermal lymphatic vessels are known to play important roles in tissue fluid homeostasis, lipid absorption, and immune surveillance (8). Of note, lymphatic vessels function as collectors and export conduits of inflammatory cells, representing gatekeepers for macrophage and lymphocyte abundance in different tissues (9). Pathological processes of inflammation, wound healing, and adipogenesis, all relevant for type 2 diabetes, have been linked to functional defects of the lymphatic system in animal experiments (9). However, for human patients, it is currently unknown whether lymphatics remain unchanged, are passive bystanders, or participate actively in the skin lesions of type 2 diabetes.

In this article, we report on enhanced lymphatic microvessel density in the skin of type 2 diabetic patients. By comparing the gene expression profiles of freshly isolated dermal lymphatic endothelial cells (LECs) from patients with type 2 diabetes with those of normoglycemic controls, we identified molecular and cellular processes regulated in lymphatic vessels, in particular, proinflammatory, lymphangiogenic, and enhanced lipid shuttling properties, accompanied by downregulated immune defense, apoptosis mediators, and small compound transporters. Concomitantly, we traced a strong dermal CD68+ macrophage infiltration, which elicited elevated tumor necrosis factor-α (TNF-α) levels. A subset of diabetic LEC (dLEC) deregulated genes was TNF-α responsive and correlated with lymphatic vessel remodeling and inflammation, including the chemokine CXCL10, which specifically led to macrophage attraction and adhesion to LECs in vitro. Hence, we have obtained the first indications to our knowledge that the dermal lymphatic system is actively involved in the progression of skin manifestations in type 2 diabetes.


Skin samples from type 2 diabetic and nondiabetic patients.

The study was approved by the local ethics committee (proposal no. 449/2001; 81/2008), and all patients (described in Supplementary Table 1) gave informed consent. Skin samples (n = 4 in each group) were taken from the proximal region of amputated legs or abdominoplastic tissue, and care was taken to excise areas at maximal distance from inflammatory or ulcerous changes (∼15 cm).

Immunohistochemical analyses.

Immunohistochemical stainings of paraffin-embedded or cryofixed skin sections were performed as described previously (10). Supplementary Table 2 summarizes the antibodies and respective dilutions applied. For quantifications, under exclusion of empty areas, nonoverlapping microscopic fields (regions of interest [ROIs]) of 100 µm2 (30 fields per patient) were captured with an Olympus VANOX AHBT3 microscope. Positively stained vessels, cellular nuclei, and macrophages were counted in these ROIs, and the cross-sectional dimension (referred to as diameter) of the vessels was measured. Average numbers were calculated per patient group and statistically analyzed as detailed later.

Ex vivo isolation of dermal LECs.

Micropreparation of LECs was performed as described previously (10). Briefly, human skin was dermatomized and epidermis and dermis dislocated by incubation in dispase solution (Roche no. 04942086001). Cells were labeled with antibodies in a three-step procedure with intermediate washing steps and sorted (FACStar Plus; BD Biosciences, Franklin Lakes, NJ) as LECs (CD31+ podoplanin positive) and blood endothelial cells (CD31+ podoplanin negative) with the use of CD45 as a gate to exclude leukocytes.

RNA isolation and chip hybridization.

GeneChip analysis was performed as described previously (10). Briefly, total RNA (RNeasy Mini Kit; Qiagen no. 74104) was amplified (MessageAmp II aRNA Amplification Kit; Ambion no. AM1751), and 1 μg of amplified RNA was biotin labeled, purified (MessageAmp II-Biotin Enhanced Kit, Ambion no. AM1791) and hybridized to Affymetrix GeneChips (GeneChip Human Genome U133 Plus 2.0 Arrays), which were scanned with GeneChip Scanner 3000 7G.

Bioinformatic data analysis.

dLEC and nondiabetic LEC (ndLEC) gene expression profiles were analyzed with GeneSpring GX software (Ambion). After background correction, raw data were normalized with the robust multiarray average method, and P values were calculated by unpaired t testing. Subsequently, 1,828 probe sets with P ≤ 0.05 were used for hierarchical clustering based on entities and conditions according to Euclidean distance metric and centroid linkage rules. All data have been deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) and are accessible through GEO accession number GSE38396 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE38396). Further in-depth bioinformatic analysis was performed by relative variance method (RVM), which allows for statistical analysis of small sample sizes based on a self-adaptive threshold (11). Extensive NCBI GEO and PubMed (http://www.ncbi.nlm.nih.gov/sites/entrez) research was amended to retrieve information on gene expressions and relevance for LEC biology. Gene ontology (www.affymetrix.com) was used to classify genes according to functionality. Ingenuity Pathway Analysis (http://www.ingenuity.com) was applied to identify associations with particular functions, canonical pathways, and diseases.

Quantitative reverse transcriptase–polymerase chain reaction.

One microgram of amplified total RNA was transcribed into cDNA by Superscript II Reverse Transcriptase (Invitrogen no. 18064-014). A complete probe list is summarized in Supplementary Table 5.

Dermal LEC cultivation, TNF-α stimulation, and CXCL10 ELISA.

LECs were sorted from human dermal microvascular endothelial cells (Promocell no. C-12260) by antipodoplanin antibody–coated magnetic beads. LECs were grown in endothelial basal medium-2 (EBM-2) supplemented with 5% FCS and EGM-2-MV SingleQuots (Lonza no. CC-4147) at 37°C and 5% CO2. For TNF-α stimulations, LECs between passages 5 and 8 grown to 70–80% confluency in six-well plates were starved overnight in EBM-2/0.5% FCS and then medium containing 10 ng/mL TNF-α (R&D no. 210-TA-010) for 6, 12, and 24 h with or without 25 ng/mL inhibitory anti-TNF-α antibody was added. LECs were lyzed with RL buffer (Qiagen no. 79216) and β-mercaptoethanol for subsequent quantitative PCR (qPCR) analysis. LEC culture supernatants were collected and concentrated 20-fold with centrifugal filter units (Amicon Ultra-0.5 10K Ultracel R; Millipore no. UFC501024). Ninety-six-well ELISA plates were coated with concentrated supernatants overnight, and CXCL10 ELISA was performed according to standard protocol (see legend of Supplementary Fig. 8B). Experiments were performed in triplicate, and statistical differences between groups were analyzed as detailed later.

LEC motility and population doubling assay.

For migration assay, a scratch wound was created in confluent LEC monolayers grown in 24-well plates. Nonadherent cells were washed away, fresh medium with or without TNF-α as described previously (n = 3) was added, and wound closure was monitored every hour with an inverted live cell microscope (Axiovert 200M; Zeiss). For population doubling analysis, identical LEC numbers were grown in six wells with or without addition of TNF-α (three in each group). After defined time points, cells were washed, trypsinized, and counted in a hemocytometer. Population doublings were calculated as ln(cell concentration counted/cell concentration seeded).

Macrophage adhesion and chemotaxis assay.

Human monocytic leukemia cell line THP-1 (American Type Culture Collection no. TIB-202) was grown in RPMI 1640/10% FCS (Gibco no. 10108)/1% Pen/Strep and labeled with CellTracker Green CMFDA (5-chloromethylfluorescein diacetate) Molecular Probes 1:5,000 (Invitrogen no. C2925). Five by 105 cells were stimulated with 10 ng/mL phorbol myristic acid (PMA) (Sigma no. P8139) for 24 h, diluted in 500 μL EBM-2/0.5% FCS, and immediately added to LEC monolayers (n = 3). In blocking experiments, LECs were preincubated for 1 h with inhibitory anti-CXCL10 and anti-vascular cell adhesion molecule 1 (VCAM-1) antibodies before addition of THP-1 cells. After 3 h, nonadherent cells were washed away, and adherent THP-1 cells were photographed with an inverted live cell microscope (Axiovert 200M). Agarose spot assay was performed as described previously (12). Briefly, 1% agarose solution (Gibco no. 18300-012) was mixed 1:1 with concentrated LEC culture supernatants (see previous). Two 10-µL spots of this mixture were pipetted into 24-well plates and allowed to gel for 10 min at 4°C. Five by 105 THP-1 cells labeled and stimulated as described previously were added to each well. Inhibitory antibodies to TNF-α and CXCL10 were added 6 h before the end of the incubation period. After 24 and 48 h, images were taken, and the area covered by chemoattracted macrophages was measured with AxioVision version 4.7 software and plotted as the ratio of covered area to area at time 0.

Statistical analysis.

Analyses were performed with Microsoft Excel 2007. Variance diversity of respective data sets was determined by the F test, and significance of difference was evaluated by Student t test. P ≤ 0.05 was considered significant.


Enhanced lymphatic vessel density in type 2 diabetic skin.

Immunohistological examination revealed thickened, laminin, and type IV collagen containing basement membranes of dermal blood vessels in type 2 diabetic patients (Fig. 1A and Supplementary Fig. 1A), which is characteristic for diabetic microangiopathy. Podoplanin-positive lymphatic vessels were devoid of laminin yet covered by type IV collagen in both groups (Fig. 1A and Supplementary Fig. 1A), indicating no changes in basement membrane composition. The average lymph vessel diameter was similar in both groups (not shown), suggesting an absence of lymphatic microvascular hyperplasia or edematous condition. Duffy antigen receptor for chemokines (13,14) was established as a distinct marker for blood vessel stainings of human skin sections (Supplementary Fig. 1B). In contrast to unaltered blood vessel counts, lymphatic vessel density was significantly increased in type 2 diabetes (1.8-fold, P = 0.035) (Fig. 1B). Moreover, whereas no Ki-67–positive nuclei were found in nondiabetic samples (Fig. 1C, arrowheads), 2% of nuclei (Fig. 1C and Supplementary Table 3) in podoplanin-positive lymphatic vessels expressed the proliferation marker Ki-67 in diabetic skin (Fig. 1C, arrows). Together, these data suggest that diabetic skin displays a higher lymphatic vessel density, which might be attributed to increased proliferation of LECs.

FIG. 1.

Lymphatic vessel morphology, density, and proliferation in the skin of type 2 diabetic vs. nondiabetic patients. A: Paraffin sections of human type 2 diabetic and nondiabetic skin (n = 4 in each group) were double stained with periodic acid Schiff (PAS) (pink) and podoplanin (brown) to visualize glycoprotein deposits in basement membranes. PAS staining was prominent around diabetic blood vessels but absent from lymphatic vessels. Frozen sections of type 2 diabetic and nondiabetic human skin were double stained with antibodies to podoplanin (green) and the ECM protein laminin-α (red). Merged images show that laminin-α was absent from lymphatic vessels. Scale bars = 20 μm. B: Representative images and quantitative evaluation of immunohistochemical stainings of paraffin-embedded skin sections with antibodies to Duffy antigen receptor for chemokines for blood vessels and podoplanin for lymphatic vessels. Nonoverlapping fields (100 µm2) of immunohistochemical stainings were captured and vessels counted within ROIs under omission of empty space. As indicated by arrows vs. arrowheads, in diabetic skin, blood vessel count was elevated 1.3-fold, but this was not significant (P = 0.7), whereas lymphatic vessel density was increased 1.75-fold (P = 0.035). Scale bars = 100 μm. *P ≤ 0.05. C: Examples of lymphatic vessels in type 2 diabetic skin coexpressing podoplanin (pink) and nuclear proliferation marker Ki-67 (brown) in LECs and their quantification. An average 2% of LEC nuclei were positive for Ki-67 in diabetic skin (arrows), whereas no Ki-67–positive nuclei (arrowheads) were found in lymphatic vessels of nondiabetic skin (P < 0.05). Scale bars = 20 μm. *P ≤ 0.05. NS, not significant.

dLEC gene expression profile.

To identify transcriptomal changes of lymphatic vessels in type 2 diabetes, we performed cDNA microarray analyses of ex vivo isolated dermal LECs (Supplementary Fig. 2) of four type 2 diabetic and four normoglycemic patients. Hierarchical clustering analysis of deregulated probe sets (P ≤ 0.05) revealed separation of dLEC and ndLEC samples into two groups (Fig. 2). Detailed statistical testing led to identification of the 160 top deregulated transcripts in dLECs, of which 38 were upregulated and 122 downregulated (Table 1). According to NCBI database screening, only 40% of the transcripts (63 genes) have been reported in LECs, lymph fluid, or both so far. Expression of established LEC identity markers was unchanged between dLECs and ndLECs (Supplementary Table 4), whereas deregulations of previously detected transcripts (10,15) indicated altered lymphatic properties in type 2 diabetes. Regulation of several genes (HHEX, TSPAN8, LMNA) (4) that were genetically linked to metabolic diseases showed that these markers prove valid in the lymphatic vessel compartment. Differential gene expression of selected gene candidates was confirmed by qPCR (Fig. 3A) and by immunofluorescence (Fig. 3B and C) at mRNA and protein levels, respectively (n = 4 per group). These findings revealed that ex vivo isolated LECs exhibit a distinct gene expression signature in type 2 diabetes.

FIG. 2.

dLECs exhibit a gene expression pattern separate from ndLECs. Hierarchical clustering of dLEC and ndLEC microarray gene expression data was based on 1,828 significantly deregulated probe sets shows grouping of dLECs and ndLECs in two separate branches. On the y-axis, an entity tree was generated by grouping the probe sets of the eight array samples according to the similarity of their expression profiles. On the x-axis, a condition tree was generated to show the relationship between the samples. The color range indicates the fold increased (red) and decreased (blue) expression difference of probe sets between dLECs and ndLECs.


Transcript levels of 160 deregulated candidate genes functionally clustered in inflammatory response, LEC adhesion and migration, growth and lymphangiogenesis, and small molecule biochemistry

FIG. 3.

Confirmation of altered gene expression in lymphatic vessels of type 2 diabetic patients on the mRNA and protein levels. A: Representative gene candidates with strong expression change in dLECs vs. ndLECs were selected, and qPCR reactions were performed in triplicate, with total RNA extracted from dLECs and ndLECs (n = 4 in each group) as a template. Fold changes of respective gene expression ratios and statistical significance thereof were calculated and plotted (gray bars) beside the microarray results (black bars) (P values derived from Table 1). *P < 0.05; **P < 0.01; ***P < 0.001. B and C: Representative images of 5-μm frozen sections of human skin double stained with antipodoplanin and respective antibody to selected candidate genes (n = 4 in each group). Confirmation of enhanced CXCL10, VCAM1, FABP4, and CYR61 (B) and of reduced CXADR and AQP3 (C) expression. Scale bars = 20 μm.

Modulation of functional transcript groups in dLECs.

To understand cellular processes associated with the dLEC transcript signature, we introduced a hierarchical categorization of the regulated genes on three levels. Gene ontology categories and manual annotation were used to assign basic functional descriptions to the genes, integrating these into distinct cellular process subgroups. Combined with Ingenuity Pathway Analysis (Supplementary Fig. 3), this led to the establishment of four overrepresented themes: 1) defense response and inflammation, 2) tissue remodeling and LEC motility, 3) lymphangiogenesis and cell fate regulation, and 4) lipid handling and small molecule biochemistry (Table 1). dLEC transcripts were highly enriched for plasma membrane (44 transcripts) and cell periphery and extracellular space (71 transcripts) (both P < 0.001, not shown), highlighting an extensive modulation at the environmental interface.

Altered defense and inflammation response of dLECs.

Within gene cluster A (Table 1), established acute phase genes (HP, PTX3) and wounding factors (NOX4, TXNL4B) were upregulated, indicating an active reaction to injury signals, whereas expression of antimicrobial and defense response genes (DEFB1), inflammatory response genes (PTGS2, PLA2G4A), and immediate early genes (CRIP1) were downmodulated. Transcripts for chemotactic factors and leukocyte trafficking receptors (CXCL10, VCAM1, CMTM7) were enhanced, whereas several chemokines (CCL27, CXCL14, STC1), transcripts of interleukin-1 antagonists (IL1R2), and of anti-inflammatory interleukin signaling receptors (IL20RB, IL13RA2) were suppressed, pointing toward reduced signaling activation probably because of an enhanced cytokine milieu. Overall, the data show induction of certain acute phase, chemotactic, and stress response genes in dLECs, whereas several factors involved in the elimination of infections and inflammation are suppressed, suggesting a hampered anti-inflammatory response in dLECs.

Altered cell motility and morphogenesis of dLECs.

Cluster B (Table 1) included a group of decreased transcripts for secreted growth factors (CGA, AREG), their carriers (FGFBP1, IGFBP3, TGFBI), and signaling components (FGFR3, ANXA8L2), indicating altered sensitivity to growth factor signals. Transcripts for extracellular matrix (ECM) degradation proteases (TLL1, MMP2) were induced, whereas their counterparts, proteinase regulators and inhibitors (SERPINs, CAPNS2), were suppressed, suggesting that dLECs actively pursue tissue degradation. Transcripts encoding integrin-binding and single-cell motility molecules (CYR61, LYVE1) were upregulated, whereas several transcripts for secreted proteoglycans (SDC1, DCN), ECM adhesion proteins (GPNMB, CD44), and filopodia formation proteins (LY6D, IFFO2, ANK3) were reduced, probably because of enhanced dLEC mobility. Well established was a group of decreased transcripts for cell-cell adhesion (CXADR, CLDN1) and desmosomal proteins (DSC3, PKP1, DSG3), pointing to altered mutual interaction between dLECs. The findings suggest that dLECs show a characteristic regulation of growth factor response, cell attachment, and adhesion transcripts, indicating overall ongoing lymphatic vessel remodeling.

Altered lymphangiogenesis and cell fate regulation of dLECs.

Cluster C (Table 1) included transcripts coding for axon guidance factors (NLGN4X, ROBO1), suggesting phenotypic dLEC changes. Guanosine triphosphate enzyme (GTPase) activators (RAPGEF2, RGS17) were enhanced, whereas several small GTPase-associated transcripts (RAB25, RND3, TAGAP) were decreased, which might point toward sustained G-protein signaling during lymphangiogenic processes. A series of transcription factors were downmodulated, including transcriptional repressors (BHLHB3, ID4), regulators of proliferation (HOPX, ID4, PER2), and metabolism and cell growth (FOXQ1, KLF5, TFAP2A), many of these yet unknown in LECs except for the upregulation of JUN and HHEX, which foster cell growth and inflammation. Furthermore, downregulation of transcripts for promitotic (TPPP3, ANLN) and proapoptotic (LGALS7B, P21, PERP) factors suggested that dLECs reside in a postmitotic state accompanied by strong survival promotion. Overall, these findings suggest reactivated growth of dLECs. Importantly, the identified gene sets are in accordance with the initially observed enhanced lymphatic vessel density.

Altered lipid and small molecule handling of dLECs.

Eminent within cluster D (Table 1) was upregulation of transendothelial lipid transport and handling molecules (CYP46A1, FABP4, APOD) and downregulation of enzymes involved in lipid oxidation and fatty acid degradation (GPR109B, SCD-1, SLC27A2), indicating increased influx and transport of free fatty acids by dLECs probably because of enhanced lipid status in the dermis of type 2 diabetic patients. Furthermore, transcripts for glucose (SLC2A1), amino acid (SLC7A2), water (AQP3), and chloride anion (CLCA2) transporters and for small compound metabolic enzymes (KYNU, CA12, CPS1) were decreased in dLECs, suggesting disturbed transport and biochemistry of small molecules.

Enhanced macrophage density and TNF-α levels in type 2 diabetic skin.

To correlate the dLEC transcriptome with phenotypic changes of the skin, we performed bioinformatic pathway analyses to show enrichment of gene signatures related to immune response, inflammation, and leukocyte trafficking (Supplementary Fig. 4). As one mechanism for diabetic skin complications, the enhanced presence of alternatively activated macrophages was proposed (16). Hence, we explored immune cell abundance in the same skin samples used for LEC isolation. Although T-cell and dendritic cell counts were not significantly altered (not shown), we found a massive 3.5-fold increase (P < 0.001) of CD68+ macrophages in type 2 diabetic skin (Fig. 4A). Growing evidence in the literature suggests insulin resistance as a result of a chronic inflammatory milieu (3) that involves macrophage-derived TNF-α as a prominent pathophysiological stimulus (17). After establishing TNF-α immunofluorescence stainings (Supplementary Fig. 5), type 2 diabetic skin revealed enhanced TNF-α abundance (Fig. 4B) and double immunofluorescence localized TNF-α expression to CD68+ macrophages (Fig. 4C). Densitometric evaluation showed that 60% of these cells expressed TNF-α in diabetic skin but only 20% in nondiabetic skin (Fig. 4D), which denoted an additional threefold increase of TNF-α levels. Macrophage infiltration is a major driver of prolymphangiogenic processes by secretion of lymphangiogenic factors (18). We determined that 90% of the macrophages expressed vascular endothelial growth factor (VEGF)-C and VEGF-A in both skin subtypes (Supplementary Fig. 6), indicating moderately enhanced levels in type 2 diabetes because of enhanced macrophage levels. Essentially, we traced elevated TNF-α levels produced by macrophages that were extensively infiltrating the skin of type 2 diabetic patients.

FIG. 4.

Enhanced TNF-α levels in human type 2 diabetic skin as a result of macrophage infiltration. A: Immunohistochemical evaluation of macrophage infiltration in the skin of type 2 diabetic versus normoglycemic patients. Dermal macrophages were stained with anti-human CD68 antibody. Positive cells were counted per field, and average numbers were calculated per patient group (n = 4 in each group). Scale bars = 100 μm. **P < 0.001. B: Representative images of immunofluorescence stainings of paraffin skin sections with anti-TNF-α antibody and DAPI as nuclear counterstaining. Although an antibody isotype control showed no reactivity, increased TNF-α staining was detected in type 2 diabetic vs. nondiabetic skin. Scale bars = 100 μm. C: Representative images of double immunofluorescence stainings of frozen sections of human type 2 diabetic and nondiabetic skin samples revealing cytoplasmic TNF-α abundance in CD68+ cells. Scale bars: = 20 μm. D: Quantitative evaluation of CD68+ macrophages colocalizing with TNF-α in nondiabetic vs. diabetic skin (n = 4 in each group) was performed as described in A. Scale bars = 100 μm. **P = 0.002.

TNF-α recapitulates dLEC gene expression changes and leads to enhanced LEC motility.

To analyze whether the dLEC transcriptome mirrored enhanced interstitial TNF-α levels, we performed an in-depth analysis of data on LECs stimulated with TNF-α in vitro (19,20) (Supplementary References 1 and 2) and identified a gene set (CXCL10, VCAM1, CYR61, AQP3, CXADR, SDC1) that matched with the dLEC expression changes. Primary human dermal LECs were controlled for stable identity in vitro (Supplementary Fig. 7). Thereafter, TNF-α treatment recapitulated upregulation of CYR61 and suppression of AQP3, CXADR, and SDC1 transcripts, similar to ex vivo dLECs (Fig. 5A). Anti-TNF-α blocking antibody abolished respective up- and downregulation of the transcripts (Fig. 5A), confirming TNF-α–dependent gene regulation. We aimed to explore whether TNF-α–modulated gene expression changes had functional relevance in LECs. CYR61 is an ECM-associated protein that stimulates angiogenesis and vascular integrity (21), CXADR is a homophilic cell adhesion molecule essential for lymph vessel development (22), and SDC1 is a proteoglycan that modulates ECM binding (19). Hence, CYR61 upregulation and CXADR and SDC1 downregulation suggested reduced intercellular adhesion and enhanced migratory behavior of LECs. Area closure of scratch wounds in LEC monolayers occurred significantly faster under TNF-α treatment (Fig. 5B), which was not a result of increased cell population doublings because TNF-α–treated cells showed significantly lower doubling rates than untreated controls (Fig. 5C). These data suggest that TNF-α–mediated gene expression changes evoke a lymphatic vessel remodeling phenotype that depends on increased LEC motility.

FIG. 5.

TNF-α induces altered gene expression involved in a promigratory LEC phenotype in vitro. A: qPCR analysis of candidate gene expression in LECs stimulated with TNF-α for 24 h (with or without TNF-α antibody) (n = 3). *P < 0.05; **P < 0.01; ***P < 0.001. B: Representative images of an LEC migration assay after creating wounds in LEC monolayers and (n = 3) grown with or without the addition of TNF-α. Reoccupation of the gap by LECs after time (t) = 12 h and t = 48 h (gray line) versus t = 0 h (black line) is seen. The areas repopulated by LECs were measured using AxioVision version 4.7 software, and the percentage of gap area newly covered by LECs was calculated at each time point; *P = 0.007. C: LEC doubling rates were reduced under addition of TNF-α. After respective time points, LECs grown in six wells with or without addition of TNF-α were washed, harvested, and counted, and the population doubling was calculated according to ln (cell concentration counted/cell concentration seeded); *P < 0.05. Data are mean ± SD (n = 3 in each group).

TNF-α stimulates macrophage adhesion to LECs.

CXCL10 and VCAM1 were the transcripts most strongly regulated by TNF-α. We confirmed TNF-α–mediated CXCL10 and VCAM1 upregulation in vitro, which was suppressed by anti-TNF-α antibody (Fig. 6A). Because both are involved in the attraction and regulation of leukocyte adhesion and migration to inflamed LECs (20,23), we sought to determine whether they were also relevant for the interaction of macrophages with LECs. TNF-α–treated LEC monolayers showed twofold-enhanced macrophage adhesion in vitro, which could be completely reversed by anti-TNF-α blocking antibody (Fig. 6B). Although an anti-VCAM-1 blocking antibody did not interfere with macrophage adhesion to LECs (Fig. 6C), inhibitory CXCL10 antibody reduced macrophage adhesion to baseline levels (P = 0.04) (Fig. 6C), indicating a role of CXCL10 for macrophage interaction with LECs.

FIG. 6.

TNF-α induces altered macrophage adhesion to LECs through CXCL10 secretion in vitro. A: Quantitative real-time PCR analysis of CXCL10 and VCAM1 expression in human dermal LECs stimulated with TNF-α for 24 h (with or without TNF-α antibody [Ab]). ***P < 0.001. B: Representative images of macrophage adhesion assays and quantitative evaluation thereof. CellTracker Green–labeled and PMA-stimulated THP-1 cells were added to LEC monolayers (n = 3) treated without or with TNF-α and with anti-TNF-α Ab. Macrophages showed enhanced adhesion to TNF-α–treated LEC monolayers, which was blocked by anti-TNF-α Ab. Scale bars = 100 μm. **P < 0.01. C: CellTracker Green–labeled and PMA-stimulated THP-1 cells were added to LEC monolayers (n = 3) treated without and with TNF-α, and with anti-VCAM-1 Ab and anti-CXCL10 Ab, and adhesion was assessed as described previously. Scale bars = 100 μm. *P < 0.05. D: CXCL10 secreted by LECs is chemotactic to macrophages. Representative images of an agarose migration assay performed in triplicate and its quantitative evaluation. Migration of CellTracker Green–labeled THP-1 cells (top) into the agarose spot (bottom) after 48 h (red line) vs. 0 h (black line) is seen. Macrophage chemotaxis was enhanced toward agarose spots containing supernatants of TNF-α–stimulated LECs and was inhibited by TNF-α–blocking and CXCL10-blocking Ab. *P < 0.05; **P < 0.01. NS, not significant.

LEC-derived CXCL10 is chemotactic for macrophages.

To analyze the contribution of LEC-derived CXCL10 to macrophage adhesion, we confirmed enhanced CXCL10 protein levels in LEC lysates and culture supernatants (Supplementary Fig. 8A and B, respectively) after TNF-α stimulation. Macrophage migration into agarose plugs containing culture supernatants of TNF-α–stimulated LECs showed a twofold increase and could be specifically blocked by inhibitory anti-TNF-α and anti-CXCL10 antibodies (Fig. 6D). These data provide evidence for TNF-α–induced CXCL10 secretion of LECs that leads to chemotactic recruitment of macrophages (Fig. 7).

FIG. 7.

A cascade of pathogenic events may lead to lymphatic vascular reorganization in type 2 diabetic skin. Increased load of metabolites and macrophage influx over time might lead to molecular alterations in dermal LECs. Hence, dLECs reveal an activated phenotype characterized by increased inflammatory, migratory, and lymphangiogenic status and apoptosis resistance. This lymph vessel phenotype may contribute to chronic inflammation, decreased defense against infections, leukocyte recruitment, tissue remodeling, and severely altered skin homeostasis. Specifically, TNF-α is a key mediator of cross talk between proinflammatory macrophages and LECs. TNF-α–mediated gene expression alterations in dLECs may lead to further macrophage recruitment, reinforcing lymph vessel expansion and chronic inflammation. RAGE, receptor of advanced glycation end product.


Little is currently known about the pathomechanisms of skin manifestations in type 2 diabetes. Here, we aimed to characterize physiological changes of dermal lymphatic vessels by using freshly isolated human LECs. Although thoracic duct and muscle lymphatic vessel alterations were described in diabetic mice (24,25), their skin phenotype has been ignored so far. Analyses of dermal lymphatic functioning, such as lymphangiography, would advance an understanding of disease pathogenesis; however, the single gene defects of these mice have raised doubts about studying impaired diabetic wound healing (26). To our knowledge, this study is among the first (27) to describe morphological and molecular alterations of human dermal lymphatic vessels in type 2 diabetes.

We observed an increased lymphatic vessel density, which was also demonstrated in patients with atherosclerotic lesions (28) and chronic venous insufficiency ulcers (29). In contrast to carcinoma-associated lymphangiogenesis (30), we could not detect laminin and type IV collagen expression alterations, which is consistent with unaltered lymphatic vessel diameter and absence of hyperplasia. We detected reactivated LEC proliferation, as seen by the positive Ki-67 signals in diabetic lymph vessels. This finding was supported by expression changes in p53 target genes and transcription factors implicated in the regulation of proliferation, which might cause a relief of the quiescent state of LECs. Accordingly, p53-deficient mice have shown enhanced lymphangiogenesis (31), and silencing of p53 signaling improved diabetic wound healing as a result of increased lymph vessel generation (32). We traced the downregulation of functionally coherent transcripts such as G-protein–coupled receptor (GPCR) signaling components, p53 effectors, and antiproliferative transcription factors. However, the master regulators dictating the concerted gene suppressions of several gene clusters remain unknown, suggesting involvement of epigenetic, transcriptional, or microRNA-mediated mechanisms. Other findings point toward such regulatory traits in diabetic endothelial cells (3335). Additionally, enhanced lymphatic vessel density might be the result of aberrant growth, producing dysfunctional lymph capillaries, and both phenomena might contribute to increased lymph vessel count in the skin of diabetic patients.

The dLEC transcriptome showed neither downregulation of key lymphatic differentiation markers as in obesity and inflammation (36,37) nor deregulation of transcripts for lymphatic valve proteins, although disrupted lymphatic vessel integrity has been postulated to contribute to obesity (38). Rather, overlaps of expression signatures with LECs under hypoxia and from skin of patients with lymphedema (39,40) emphasize a role of these genes in pathologic lymphatic dysfunction. Furthermore, we observed an overlap with transcriptomes of diabetes complications, such as whole-tissue lysates from nonhealing venous ulcers (41), wound inflammation (42), and diabetic wound microbiome (43), indicating a common diabetes-related gene signature.

Although macrophages are essential for physiological wound healing (44), their increased influx has been correlated with impaired wound healing in type 2 diabetic mice (45), which is characterized by expression of proinflammatory cytokines, especially TNF-α (16). Here, we demonstrate that infiltrating dermal macrophages are a source of TNF-α in human patients as well. We also traced some TNF-α expression in keratinocytes, which has been documented by others (46), but macrophages are supposed to be its predominant source (47). Activated macrophages directly contribute to de novo lymphangiogenesis (18) by 1) production of lymphangiogenic factors and 2) conversion into LECs. Pathologic lymphangiogenesis might be driven by additional mediators, such as platelet-derived growth factor-BB, hypoxia-inducible factor, and fibroblast growth factor (FGF) 2 (9), although we did not detect altered expression of these factors or their cognate receptors in dLECs. Rather, we identified a TNF-α responsive set of genes that could be confirmed in vitro, thus excluding indirect effects as a result of LEC cultivation before analysis (10). The TNF-α–induced gene deregulation in human LECs was strongly correlated with a promigratory LEC phenotype, which is in line with enhanced TNF-α levels as a driver of lymph vessel remodeling (48).

In humans, enhanced lymph vessel density correlated with positive disease outcomes (49,50). However, several reports indicated that inflammatory-driven lymphangiogenesis leads to dysfunctional vessel formation with reduced lymph drainage capacity (9). Macrophage recruitment to dLECs through CXCL10 chemotaxis might be indicative of such impeded clearance but could also be a prerequisite to lymph vessel integration. In a lipopolysaccharide-driven peritonitis model, increased numbers of macrophages were closely attached to newly formed inflammatory lymphatic vessels, directly incorporating into these (51). It is tempting to speculate that human skin macrophages are involved in an analog manner of lymph vessel expansion. Aberrantly grown lymphatic vessels could lead to decreased tissue fluid, lipid and immune cell drainage, and finally, persistent inflammation that all together hinder skin regeneration. The question of whether exaggerated lymph vessel formation is beneficial for type 2 diabetes skin pathology must be explored.


M.H. and T.K. were funded within the PhD-CCHD (Cell Communication in Health and Disease) program (http://www.meduniwien.ac.at/phd-cchd) supported by the Austrian Science Fund (FWF project no. W1205). G.E. was supported by an Elise Richter Fellowship (FWF V102-B12) and an FP7 Marie Curie International Reintegration grant (IRG230984).

No potential conflicts of interest relevant to this article were reported.

M.H. performed research, analyzed data, and wrote the manuscript. T.K. performed research and analyzed data. G.E. researched data and reviewed and edited the manuscript. H.S., C.W.S., and M.K.B. performed research. D.S. researched data. C.N. provided human skin material and clinical data and contributed to the discussion. D.K. conceived the project and reviewed and edited the manuscript. B.H. designed the research, analyzed data, and wrote the manuscript. B.H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Dr. Daniela Haluza, Institute of Environmental Health, Medical University of Vienna (MUW), and Romana Kalt and Ingrid Raab, Clinical Institute of Pathology, MUW, for excellent technical assistance. The authors also thank Dr. Martin Bilban, Department of Laboratory Medicine, MUW, for help with biostatistics; and Dr. Maximilian Zeyda, Department of Internal Medicine, MUW, and Dr. Andrew Rees and Dr. Georg Krupitza, Clinical Institute of Pathology, MUW, for helpful discussions.

  • Received June 22, 2012.
  • Accepted February 9, 2013.

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. See http://creativecommons.org/licenses/by-nc-nd/3.0/ for details.


| Table of Contents

This Article

  1. Diabetes vol. 62 no. 7 2509-2529
  1. Supplementary Data
  2. All Versions of this Article:
    1. db12-0844v1
    2. 62/7/2509 most recent