signature=26a929c349c7427d016ad99ebfc3f83f,Galectin-3 Induces a Pro-degradative/inflammatory Gene Si...

Gal-3 Localization in Clinical Specimens and OA Chondrocytes

Having collected clinical material of OA articular cartilage and subchondral bone that covers the range of the Mankin score (MS), the measurement of cellular positivity at different levels of degeneration was possible. As exemplarily shown in Fig. 1a for mildly (MS2), moderately (MS6) and severely (MS10) degenerated regions, the percentage of Gal-3 positivity in the chondrocyte population increased steadily. As presented in scatterplots for each patient (Fig. 1b, left) and for the entire group (Fig. 1b, right), onset and progression of the disease were positively correlated with this parameter (p 

Figure 1

Gal-3 presence in articular chondrocytes correlates with cartilage degeneration.

Histological sections of OA articular cartilage and subchondral bone (n = 13 patients) were stained with Safranin O (SO). From each patient, 3–9 regions of interest were graded using the Mankin score (MS). Consecutive sections were processed immunohistochemically and the percentages of Gal-3-positive chondrocytes were assessed at 40× and 400× magnification. Overview images (40×) were photomerged using Adobe Photoshop from single photographs. (a) Shown are exemplary series from mildly (MS2), moderately (MS6) and severely (MS10) degenerated cartilage regions stained with SO (left) and using an antibody against Gal-3 (middle, scale bar: 500 μm, squares indicate the regions illustrated at 400× magnification; right, scale bar: 50 μm). (b) Shown are scatterplots of Mankin scores versus percentages of Gal-3-positive chondrocytes with regression lines for each patient (left panel) and over all patients (right panel). Pearson correlation coefficients ranged from 0.43 to 0.94 (mean: 0.69 ± 0.18).

Extending our findings from immunohistochemical experiments, Gal-3 presence was visualized in OA chondrons using immunofluorescence and single plane microscopy (Fig. 2a). Z Stack projection identified Gal-3 to be consistently present throughout the cytoplasm of chondrocytes (Fig. 2a, right). When applying FITC-labelled Gal-3 to the fixed material as sensor for accessible binding sites, staining was obtained (Fig. 2b). It was less homogeneously distributed within the cytoplasm than the signal for Gal-3 presence. In fact, intracellular sites with reactivity for Gal-3 appeared to be restricted to defined cytoplasmic regions closely associated to the nucleus (Fig. 2b, right). Evidently, Gal-3 is present in chondrocytes in situ, and the detected correlation to the level of degeneration suggests that the lectin qualifies as effector in OA progression. To enable its activity in an auto-/paracrine manner, secretion of Gal-3 by chondrocytes is a prerequisite.

Figure 2

Localization of Gal-3 and its binding sites in OA chondrons and lactose-inhibitable binding of galectins to chondrocytes in vitro.

(a) OA cartilage sections were processed with antibodies against Gal-3 (red) – followed by immunofluorescence detection using AlexaFluor555-labelled second-step antibodies – together with DAPI (blue) prior to analysis using laser scanning microscopy. Differential interference contrast (DIC) imaging was included. Scale bar: 20 μm. A series of eight images was recorded at 1 μm intervals to create a stack in the Z axis. Shown is the projection from the Z stack generated using ZEN software. (b) OA cartilage sections were processed with Gal-3-FITC (green) together with DAPI (blue) prior to analysis using laser scanning microscopy. DIC imaging was included. Scale bar: 20 μm. A stack in the Z axis was created, based on eight serial microphotographs. Shown is the projection from the Z stack established using ZEN software. (c) Cultured OA chondrocytes were trypsinised and resuspended prior to labelling with Alexa-Fluor555-labelled Gal-3 (Gal-3-555) (red) and DAPI (blue) at 4 °C for 10 minutes in presence or absence of 0.1 M β-lactose. After 10 minutes of incubation, cells were washed and analysed using laser-scanning microscopy, with the focus plane set to the centre of cells. Scale bars: 20 μm. (d,e) Cultured OA chondrocytes were trypsinised and resuspended prior to labelling with (d) Gal-3-555 (red) and AlexaFluor488-labelled (Gal-3tr-488) (green) or (e) Gal-3-555 (red) and Gal-1-488 (green) at 4 °C for 10 minutes. After 10 minutes of incubation, cells were washed and analysed using laser-scanning microscopy, with the focus plane set to the centre of cells. Co-localization is presented in separate and overlay images. Arrows mark the staining of cell nuclei of chondrocytes, whose cell membranes have lost their integrity, by (d,e) Gal-3-555 (red), but not by (d) Gal-3tr-488 (green) or (e) Gal-1-488 (green). Shown are the results from chondrocytes of one patient, representative for three independent experiments (n = 3 patients). Scale bars: 20 μm.

Tested in medium of OA chondrocytes, 0.11 ± 0.17 ng/ml (n = 7 patients) were detected in supernatants after 24 h of monolayer culture. This value tends to be smaller (p = 0.08, paired t-test) than the 1.11 ± 1.74 ng/ml (n = 7) found in cell cultures from the same patients under identical conditions for Gal-1. Comparable to the situation for Gal-1Supplementary Fig. S2). Also, their presence in the medium did not affect the level of LGALS3 mRNA in OA and non-OA chondrocytes, which excludes an influence of these cytokines on transcription of the Gal-3 gene (Supplementary Fig. S2). Because these key constituents of a pro-inflammatory environment fail to affect Gal-3 expression and secretion, we hypothesized that the lectin is an upstream effector, via carbohydrate-dependent cell surface binding. Indeed, fluorescent Gal-3 binds to viable chondrocyte surfaces in a lactose-inhibitable manner (Fig. 2c). When having access to the cell interior, the lectin is also reactive with nuclei (Fig. 2d, left). Removing the N-terminal tail of Gal-3 results in a product called truncated Gal-3 (Gal-3tr), which lacks this reactivity (Fig. 2d, center and right). Taken together, Gal-3 shares the capacity for lactose-inhibitable cell surface binding with Gal-1, as shown by two-colour staining in Fig. 2e. This series of results prompted us to examine whether addition of Gal-3 to the medium of cultured OA chondrocytes engenders changes in the level of mRNA or secretion of disease markers.

Gal-3 Enhances Expression of Functional Disease Markers

Supplementation of medium with LPS-free Gal-3 led to a concentration-dependent effect on gene expression (upregulation of IL1B, TNFA, MMP1, MMP3 and MMP13; downregulation of COL2A1 and ACAN; Fig. 3a–e and i,j, respectively) and on the level of secretion in the cases of matrix metalloproteinases (MMPs) (Fig. 3f–h). This effect was sensitive to the presence of the cognate sugar (Fig. 3k,l) which also serves as an additional control for the absence of LPS activity. Interestingly, despite its difference in protein design to homodimeric Gal-1, Gal-3 was also effective to elicit enhanced expression of disease markers. In order to pinpoint structural properties relevant for this activity, we engineered and tested variants of Gal-3 by reducing the length of the N-terminal tail, shown in Fig. 4a. Physiologically, cleavage sites for MMPs predestine the tail for proteolytic shortening, if not protected by aggregation on cell surfaces (Fig. 4a). The fully truncated form has apparently lost its elicitor activity, and removal of one repeat in the central region can be critical (Fig. 4b,c). The integrity of the second half of this repeat region appears necessary to modulate chondrocyte gene expression for the tested disease markers (IL1B, CXCL8). The results for these two selected cases encouraged us to answer the question on the dimension of Gal-3’s impact on regulation of gene expression by genome-wide microarray in combination with molecular analyses.

Figure 3

Gal-3 regulates functional OA markers via lactose-inhibitable binding.

Chondrocytes of five OA patients were starved overnight prior to treatment with 1–50 μg/ml Gal-3 for 24h. Total RNA was isolated and relative mRNA levels of IL1B (a), TNFA (b), MMP1 (c), MMP3 (d), MMP13 (e), COL2A1 (i), and ACAN (j) were determined using RT-qPCR. Results are expressed as relative quantities (mean ± SD) with respect to untreated controls set to 1 (*p 

Figure 4

Structure-function relationship between Gal-3 and its activity in chondrocytes.

(a) Structural organization of the trimodular design of human Gal-3, together with cleavage sites for MMPs, from ref. 45 with modifications. (b,c) Primary OA chondrocytes (n = 3 patients) were treated with full-length Gal-3, Gal-3 IV-IX, Gal-3 V-IX or Gal-3tr for 24 h. The concentrations of the variants were set to obtain equimolarity (with respect to CRD) for stimulation experiments. In addition, Gal-3tr was also applied at 50 μg/ml to examine dose-response effects. RT-qPCR assays for IL1B (b) and IL8 for CXCL8 (c) were performed. Results are expressed as percent (mean ± SD) of the activity of full-length Gal-3 (#p 

Gal-3 Reprograms Gene Expression

Cell surface binding of Gal-3 to human OA chondrocytes caused a broad-scale change in their profile of gene expression. This change was fairly consistent in the four tested cell populations, as shown in Fig. 5a and b for the most up- and downregulated genes. Marked increases concerned a group of chemokine ligands (most notably CXCL8), the pro-inflammatory inducible nitric oxide synthase 2 (NOS2) and E-selectin (SELE, the docking site for leukocytes on inflamed endothelium). Independent RT-qPCR experiments for these mRNAs confirmed the array data (Fig. 5c). In addition, Supplementary Table S1 provides selected subsets of OA-related genes and glycogenes regulated by Gal-3. Gene Set Enrichment Analysis (GSEA) against C2 molecular signatures databases was performed to disclose functional implications. It revealed solid evidence for enhanced inflammation (Supplementary Fig. S3a–d). Metacore analysis of the Gal-3-upregulated gene sets disclosed connections with immune response and inflammation as well as with connective tissue, as well as autoimmune and arthritic joint diseases (Supplementary Fig. S3e–g). Using the C3 transcription factor targets (TFT) database for GSEA to delineate regulation of genes with common binding sites for transcription factors, an enrichment of genes harbouring NF-κB binding sites was unveiled (Fig. 6a). As graphically depicted in Fig. 6b, a significant proportion of the genes under the control of Gal-3 harbour such sites. Taken together, the findings of these bioinformatic analyses intimated an active involvement of NF-κB in the Gal-3-mediated upregulation of gene expression.

Figure 5

The most up- and downregulated transcripts in microarray analysis of Gal-3-stimulated chondrocytes.

(a) OA chondrocytes (n = 4 patients; numbered with “1–4”) were starved overnight prior to treatment with 50 μg/ml Gal-3 for 24 h. Following microarray analysis, heat maps of the 20 most upregulated genes and the 20 most downregulated genes were plotted. (b) For these genes, the fold-changes of mRNA levels in Gal-3-treated vs untreated chondrocytes across all four patients were calculated. The p-values are also given. (c) The results of the microarray experiments were ascertained using RT-qPCR in the same RNA samples as used in the microarray analysis. Values are given as fold-changes with respect to untreated controls set to 1. Means and SDs were computed, normality of the data was analysed using Shapiro-Wilk test, and statistical significance of the data was examined using paired t test or Wilcoxon signed-rank test. In addition, multiple testing correction based on false discovery rate (FDR) was applied to adjust the raw p-values.

Figure 6

The activity of Gal-3 in chondrocytes is mediated via NF-κB signalling.

(a) Top 10 results of GSEA analysis against the C3 (transcription factor target genes (TFT)) database in order of significance. K: number of genes harbouring the respective TFBS motif. k: number of overlapping genes induced by Gal-3. (b) Cytoscape and Enrichment Map visualization of significant C3-GSEA TFT results (FDR 

The emerging concept of NF-κB activation and participation was next experimentally verified. Upon stimulation of OA chondrocytes with Gal-3, phosphorylation of p65 was transiently increased (Fig. 6c,d). When determining the level of mRNA from IL1B gene transcription in the presence of three site-specific NF-κB inhibitors (IKK inhibitor VII, Bay-117082, CAPE), marked and dose-dependent blocking of the positive response by more than 75% was seen in each case (Fig. 6e–g). These data support the participation of NF-κB in the upregulation of transcription of Gal-3-responsive genes. Hereby, they establish a mechanistic similarity to the activity of Gal-1 in OA chondrocytes

Gal-3 Cooperates with Gal-1 to Enhance Effector Expression for Degradation

Comparing the gene sets, which were upregulated (≥2-fold, p 

Figure 7

Bioinformatic comparison of the gene sets regulated by Gal-3 and Gal-1.

(a–d) Pathway enrichment results using MetaCore’s “Compare Experiments” workflow obtained with genes consistently upregulated in both Gal-3-treated chondrocytes and Gal-1-treated chondrocytes (≥2-fold, p 

Having described these similar response profiles on the level of transcriptional regulation, it was tempting to test the hypothesis of a functional cooperation. As illustrated in Fig. 8a for IL1B gene expression, Gal-1 was more active as elicitor than Gal-3, when assayed in parallel in OA chondrocyte cultures. The same finding was further extended for MMP13 (data not shown). Considering the differences in slope in regression analysis for the correlation of chondrocyte positivity and cartilage degradation in clinical specimens, Gal-1 is present at a higher concentration in situ than is the case for Gal-3 (as shown in ref. 15 and Fig. 1, respectively). In the course of degeneration, these concentrations likely are at a sub-saturating level. Thus, we set the value of Gal-1 at 5 μg/ml and added Gal-3 in different quantities, especially at low Gal-3 concentrations of 0.1 and 1 μg/ml. Thereby, the assays were designed to reflect pathophysiological conditions. As shown in Fig. 8b, a significant increase by the addition of Gal-3 occurred in both cases. When tested at a saturating Gal-1 concentration (50 μg/ml), no enhancement by Gal-3 was expectedly seen (not shown). Notably, we have herewith discovered a functional cooperation between these two endogenous lectins under conditions of progressing cartilage degeneration.

Figure 8

Functional cooperation between Gal-3 and Gal-1 activities in chondrocytes.

(a) Chondrocytes of five OA patients were starved overnight prior to treatment with 10–50 μg/ml Gal-3 or 10–50 μg/ml Gal-1 for 24 h. Total RNA was isolated and mRNA levels of IL1B were determined using RT-qPCR. Results are expressed as relative quantities (mean ± SD) with respect to untreated controls set to 1 (*p