Characterization of a Novel Recombinant Hyaluronan Binding Protein for Tissue Hyaluronan Detection

Abstract

Tumor necrosis factor-Stimulated Gene 6 protein (TSG-6) is a hyaluronan (HA)-binding glycoprotein containing an HA-binding Link module. Because of its well-defined structure, HA binding properties and small size, TSG-6 is an excellent candidate as an alternative to animal-derived HA-binding protein (HABP) for the detection of HA. The present work describes the generation and characterization of a novel recombinant HA-binding probe obtained by fusion of a modified TSG-6 Link module with mutationally inactivated heparin-binding sequence and the Fc portion of human IgG1 (TSG-6-ΔHep-Fc) for tissue HA detection in histological samples. Direct binding assays indicated strong binding of TSG-6-ΔHep-Fc to HA, with little residual binding to heparin. Histolocalization of HA in formalin-fixed, paraffin-embedded tissue sections using biotin-TSG-6-ΔHep-Fc resulted in hyaluronidase-sensitive staining patterns similar to those obtained with biotin-HABP, but with improved sensitivity. HA was detected in many human tissues, and was most abundant in soft connective tissues such as the skin dermis and the stroma of various glands. Digital image analysis revealed a linear correlation between biotin-HABP and biotin-TSG-6-ΔHep-Fc staining intensity in a subset of normal and malignant human tissues. These results demonstrate that TSG-6-ΔHep-Fc is a sensitive and specific probe for the detection of HA by histological methods.

Keywords

Tumor necrosis factor-Stimulated Gene 6 protein glycosaminoglycans hyaluronan hyaluronan-binding protein tissue localization histological detection

Introduction

Hyaluronan (HA) is a linear glycosaminoglycan (GAG) composed of repeating disaccharide units of glucuronic acid β1-3/N-acetylglucosamine β1-4. HA is ubiquitously distributed in the extracellular and pericellular matrices of tissues and can sometimes be found inside cells (Evanko and Wight 1999; Hascall et al. 2004; Toole 2004).

In mammals, the dynamic turnover of HA is a finely tuned equilibrium of biosynthesis by three membrane-bound HA synthases and catabolism by several hyaluronidases (Weigel et al. 1997; Itano and Kimata 2002; Stern et al. 2005; Jadin et al. 2012). This controlled balance is frequently disturbed in malignancies, in which up-regulation of HA production has been associated with more aggressive and invasive behavior (Itano et al. 2008; Toole 2009; Kultti et al. 2012). Therefore, targeting tumor-associated HA has generated promising results as a new approach to cancer therapy. A pegylated recombinant human hyaluronidase (PEGPH20) has been shown to deplete tumor HA and lead to increased tumor perfusion, thereby enhancing the activity of anticancer agents in preclinical models (Thompson et al. 2010; Provenzano et al. 2012; Jacobetz et al. 2013). With the current interest towards personalized medicine to customize cancer treatment according to specific molecular features of individual tumors identified by companion diagnostics, it has become essential to evaluate the HA content of tumors to select patients who are more likely to respond to HA-depleting therapy.

Because HA is identical across tissues and species and therefore poorly immunogenic, the development of specific anti-HA antibodies has been challenging. Methods for the detection and measurement of HA levels in tissues are limited and primarily rely on cartilage-derived HA-binding protein (HABP) (Tengblad 1979; Tammi et al. 1988; Wells et al. 1991; Lindqvist et al. 1992; Lin et al. 1997; Kobayashi et al. 1999; Sakai et al. 2000; Baier et al. 2007), although recombinant versican G1 domain has also been used (Clark et al. 2011). However, the heterogeneity and lack of purity of these mixtures render them impractical for precise clinical applications such as companion diagnostic devices. Therefore, it is essential to design alternative methods and construct probes suitable for HA detection in clinical specimens.

TNF-stimulated gene 6 (TSG-6) is an ~30 kDa secreted HA-binding glycoprotein encoded by Tumor necrosis factor-Stimulated Gene 6 expressed in many different types of cells and tissues in response to a wide variety of cytokines and growth factors (Milner and Day 2003). TSG-6 plays critical roles in the formation and remodeling of HA-rich pericellular and extracellular matrices via HA-crosslinking during inflammatory and inflammation-like processes (Fülöp et al. 2003; Milner et al. 2006; Selbi et al. 2006; Simpson et al. 2009). TSG-6 is composed of a 100 amino-acid-long N-terminal HA-binding Link module and a C-terminal CUB (complement C1r/C1s, Uegf, Bmp1) module that binds to fibronectin (Lee et al. 1992; Kohda et al. 1996; Kuznetsova et al. 2008). The high affinity of the TSG-6 HA-binding Link module for HA (Kahmann et al. 2000; Lesley et al. 2002) makes it an excellent starting point to engineer a homogenous and specific reagent to detect HA.

The present study describes the engineering, mammalian cell expression and biochemical characterization of a recombinant biotinylated HA-binding probe constituted of a modified TSG-6 Link module with a mutationally-inactivated heparin-binding domain fused with the Fc region of human IgG1 (biotin-TSG-6-ΔHep-Fc). Biotin-TSG-6-ΔHep-Fc labeling was compared with that obtained with biotinylated bovine cartilage-derived HABP in formalin-fixed paraffin-embedded (FFPE) normal and neoplastic human tissues to ascertain its value as a reliable probe for the histological detection of HA in the context of companion diagnostic device development.

Materials & Methods

Cell Line and Culture Conditions

For recombinant protein expression, FreeStyle™ Chinese hamster ovary CHO-S cells were cultured in suspension in shake flasks in FreeStyle™ CHO Expression Medium from Life Technologies (Grand Island, NY) supplemented with 8 mM L-glutaMax (Life Technologies) at 37C in 8% CO₂ in air on an orbital shaker platform rotating at 125 rpm with vented caps on flasks to allow for aeration.

Vector Construction, Transient Expression and Purification of Recombinant Human TSG-6 Link Module-IgG Fc Fusion Proteins

A DNA fragment encoding human IgG1 heavy chain (GI No. 5031409), human TSG-6 Link module (GI No. 315139000) and human immunoglobulin light chain kappa (κ) leader (Giudicelli et al. 2005) (TSG-6-Fc), with unique NheI and BamHI endonuclease cleavage sites at each end, was synthesized de novo at GenScript (Piscataway, NJ) and inserted into HZ24, a Halozyme proprietary mammalian expression vector (Baker and Bookbinder 2011). A similar construct, containing a mutated version of TSG-6-Fc (TSG-6-ΔHep-Fc), was produced by replacing three lysines, which are known to be critical for heparin binding, with alanine residues at positions 55, 69 and 76 (Mahoney et al. 2005).

Transient transfection of both constructs into FreeStyle™ CHO-S cells was performed using FreeStyle™ Max lipid (Life Technologies), according to the manufacturer’s instructions.

TSG-6 recombinant proteins were affinity purified using Protein A (Bio-Rad, Hercules, CA), according to the manufacturer’s instructions.

SDS-PAGE and Western Blot Analyses

The purity, size and identity of the recombinant proteins were determined by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) using 4–20% Tris-glycine gels (Life Technologies) and western blot analysis. After electrophoresis, proteins were transferred to polyvinylidene difluoride (PVDF) membranes using iBlot® (Life Technologies). The membranes were blocked with 5% non-fat milk solution (in PBS/0.1% Tween-20), and probed with a goat anti-TSG-6 IgG (R&D Systems, Minneapolis, MN) followed by HRP-rabbit anti-Goat IgG for detection (Calbiochem/EMD Millipore, Billerica, MA) or with an HRP-rabbit anti-human IgG Fc (Jackson Immunoresearch, West Grove, PA). Images were taken using a Chemi Doc™ MP Imaging System (Bio-Rad, Hercules, CA). Full length TSG-6 (R&D Systems, Minneapolis, MN) was used as a control.

Assessment of the Direct Binding of TSG-6-Fc and TSG-6-ΔHep-Fc to HA and Heparin

The binding of TSG-6-Fc and TSG-6-ΔHep-Fc to HA and heparin was assessed using HA or heparin-coated microplates. Immulon 4HBX 96-well plates (Thermo Fisher Scientific, Waltham, MA) were coated overnight at 4C with 100 µg/ml of HA (1000 kDa average molecular weight, Lifecore, Chaska, MN) or heparin (15 kDa average molecular weight, EMD Millipore, San Diego, CA) in 0.5 M sodium carbonate buffer, pH 9.6. The binding of TSG-6-Fc or TSG-6-ΔHep-Fc was detected using an HRP-conjugated anti-human IgG Fc antibody followed by 3,3’, 5,5’-tetramethylbenzidine (TMB) substrate (Kirkegaard & Perry Laboratories [KPL], Gaithersburg, MD). One-phase association curves were fitted and the apparent Kd determined using GraphPad Prism v5 (GraphPad Software, La Jolla, CA).

Biotinylation of TSG-6-Fc and TSG-6-ΔHep-Fc

Random biotinylation of TSG-6-Fc and TSG-6-ΔHep-Fc on primary amine groups was performed using NHS-PEG4-biotin (Thermo Fisher Scientific), according to the manufacturer’s instructions.

Competitive Inhibition of Biotin-TSG-6-ΔHep-Fc and Biotin-HABP Binding to Immobilized HA using HA, Chondroitin Sulfate A, Chondroitin Sulfate C and Heparin Solutions

Immulon 4HBX 96-well plates (Thermo Fisher Scientific) were coated with 1.2 MDa HA (Lifecore Biomedical, Chaska, MN) in 0.5 M sodium carbonate buffer, pH 9.6. Biotin-TSG-6-ΔHep-Fc or biotin-HABP (Seikagaku Corporation, Tokyo, Japan) at 200 ng/ml was pre-incubated with increasing concentrations of 150 kDa HA (Lifecore Biomedical), 50 kDa chondroitin sulfate A (Calbiochem), 40-80 kDa chondroitin sulfate C (Seikagaku) or 15 kDa heparin (Calbiochem), and added to the 1.2 MDa HA-coated plates for 60 min. Bound biotin-TSG-6-ΔHep-Fc or biotin-HABP was detected using streptavidin-HRP (Jackson Immunoresearch) followed by TMB substrate (KPL). Non-linear regression in GraphPad Prism was used to calculate and plot four-parameter inhibition curves.

Competitive Inhibition of Biotin-TSG-6-ΔHep-Fc and Biotin-HABP Binding to Immobilized HA using HA Solutions of Different Molecular Weights

Immulon 4HBX 96-well plates (Thermo Fisher Scientific) were coated with 1.2 MDa HA (Lifecore Biomedical, Chaska, MN) in 0.5 M sodium carbonate buffer, pH 9.6. Biotin-TSG-6-ΔHep-Fc or biotin-HABP (EMD Millipore, San Diego, CA) at 200 ng/ml were pre-incubated with increasing concentrations of several HA preparations with different average molecular weights (Lifecore Biomedical) for 5 min and added to the 1.2 MDa HA-coated plates for 60 min. Bound biotin-TSG-6-ΔHep-Fc or biotin-HABP was detected using streptavidin-HRP (Jackson Immunoresearch) followed by TMB substrate (KPL). Non-linear regression in GraphPad Prism was used to calculate and plot four-parameter inhibition curves.

Histological Staining of HA

FFPE normal human skin was purchased from OriGene (Rockville, MD). FFPE normal and malignant human tissue microarray (TMA) slides were purchased from US Biomax (Rockville, MD). Tissue sections were labeled with either 0.04–1 µg/ml biotin-HABP (Seikagaku Corporation) or 0.04–1 µg/ml biotin-TSG-6-ΔHep-Fc overnight at 4C. Detection was performed using streptavidin-HRP (BD Pharmingen™, San Jose, CA) followed by DAB+ (Dako A/S, Glostrup, Denmark). Slides were counterstained with Gill’s hematoxylin. For specificity control, tissue sections were treated with 50 U/ml Streptomyces hyalurolyticus hyaluronidase in 20 mM sodium acetate, 75 mM sodium chloride, pH 6, for 5 hr at 37C, or with recombinant human PH20 (rHuPH20) in 25 mM piperazine-N,N’-bis(2-ethanesulfonic acid) (PIPES), pH 5.5, 70 mM sodium chloride, 0.1 % BSA, for 2 hr at 37C. Test sections were incubated in 25 mM PIPES, pH 5.5, 70 mM sodium chloride, 0.1 % BSA for 2 hr at 37C or in 20 mM sodium acetate, 75 mM sodium chloride, pH 6, for 5 hr at 37C with no enzyme. Negative controls, in which the primary reagent (biotin-HABP or biotin-TSG-6-ΔHep-Fc) was omitted, were also performed.

Photomicrography and Digital Image Analysis

For normal human skin sections, micrographs were captured at 20× and 40× objective magnifications using an Axioskop™ microscope (Carl Zeiss Microscopy, Oberkochen, Germany) equipped with a SPOT™ Pursuit digital camera (Diagnostic Instruments, Sterling Heights, MI). For TMAs, slides were scanned at 20× magnification and each tissue core analyzed using the Aperio Positive Pixel Count v9 algorithm (Aperio, Vista, CA). Positivity was measured as the number of pixels stained above a defined intensity threshold divided by the total number of pixels stained and counterstained, excluding white areas (Jiang et al. 2012). Correlation analysis between biotin-HABP and biotin-TSG-6-ΔHep-Fc staining positivity for each tissue core was performed using the GraphPad Prism v5 software.

Results

Design, Expression and Purification of Recombinant TSG-6-Fc and TSG-6-ΔHep-Fc

Recombinant TSG-6-Fc and TSG-6-ΔHep-Fc were transiently expressed in Freestyle™ CHO-S cells using the cytomegalovirus (CMV)-driven mammalian expression vector (Fig. 1A). Both recombinant proteins were purified from the culture media through a single step protein A affinity column with a yield of 3–5 mg/liter. SDS-PAGE analysis revealed a single protein species of about 40 kDa for both TSG-6-derived proteins under reducing conditions, and of about 80 kDa under non-reducing conditions, indicating the probable formation of homodimers via disulfide bonds present in the hinge region of IgG Fc (Fig. 1B). The purified proteins were stable in PBS for at least one month at 4C without any visible degradation (data not shown). Treatment with PNGase F glycosidase, which removes N-linked glycans, induced a 5-kDa shift in molecular weight for both recombinant proteins, indicating that they were N-glycosylated (data not shown). The protein identity of both TSG-6-Fc and TSG-6-ΔHep-Fc was confirmed by tryptic digestion/LC-MS analysis of the final product (data not shown) and western blot analysis using an anti-TSG-6 antibody (Fig. 1C). Confirming the results of the SDS-PAGE analysis, a 40-kDa band was also detected under reducing conditions, and an 80-kDa band was detected under non-reducing conditions. In addition, a faint band of 160 kDa was observed, probably corresponding to aggregation dimers (Fig. 1C, lanes 5 and 6). Western blot analysis using an anti-human IgG Fc yielded similar results: a 40-kDa band under reducing conditions, and an 80-kDa band under non-reducing conditions (Fig. 1D).

Figure 1.

Vector design, transient expression and purification of recombinant human TSG-6-Fc and TSG-6-ΔHep-Fc. (A) Schematic representation of the pHZ24-TSG-6-Fc expression plasmid. CMV, Cytomegalovirus promoter; SP, signal peptide; TSG-6-LM, TSG-6 Link module; hIgG1Fc, human IgG1 Fc region; solid line, plasmid backbone. (B) SDS-PAGE analysis of full length recombinant human TSG-6 used as positive control (lanes 1 and 4), CHO-S cell medium-purified TSG-6-Fc (lanes 2 and 5) and TSG-6-ΔHep-Fc (lanes 3 and 6) under reducing (lanes 1-3) and non-reducing (lanes 4-6) conditions. One µg of each protein was loaded in each lane. (C) Western blot analysis of full length recombinant human TSG-6 (used as positive control, lanes 1 and 4), TSG-6-Fc (lanes 2 and 5) and TSG-6-ΔHep-Fc (lanes 3 and 6) using a goat anti-TSG-6 IgG under reducing (lanes 1-3) and non-reducing (lanes 4-6) conditions. Sixty ng of each protein was loaded in each lane. (D) Western blot analysis of TSG-6-Fc (lanes 1 and 3) and TSG-6-ΔHep-Fc (lanes 2 and 4) using an anti-IgG Fc antibody under reducing (lanes 1-2) and non-reducing (lanes 3-4) conditions. Sixty ng of each protein was loaded in each lane. M, protein molecular weight standard.

Binding of Recombinant TSG-6-Fc and TSG-6-ΔHep-Fc to HA and Heparin

Binding of TSG-6-Fc and TSG-6-ΔHep-Fc to HA and heparin was assessed using a direct ELISA-like assay. Both recombinant proteins displayed similar HA binding, with high affinity as evidenced by the apparent Kds of 0.13 nM (9.98 ng/ml) and 0.15 nM (11.21 ng/ml) for TSG-6-Fc and TSG-6-ΔHep-Fc, respectively, and overlapping HA binding titration curves (Fig. 2, left panel). This indicates that the mutations introduced in the heparin binding site do not affect HA binding. However, as expected, there was a significant difference between TSG-6-Fc and TSG-6-ΔHep-Fc regarding their binding to heparin, with apparent Kds of 0.44 nM (33.26 ng/ml) and 1.67 nM (125.8 ng/ml), respectively (Fig. 2, right panel).

Figure 2.

Comparison of TSG-6-Fc and TSG-6-ΔHep-Fc binding to immobilized HA and heparin. Binding of increasing concentrations of TSG-6-Fc (black circles) and TSG-6-ΔHep-Fc (grey squares) to HA-coated microplates (left panel) and to heparin-coated microplates (right panel). Results are expressed in Absorbance Units (AU) at 450 nm. Each datapoint is the average of duplicate assays with SEM. Curves were fitted to one-phase association equations using GraphPad Prism.

Assessment of the Potential Cross-reactivity of Biotin-TSG-6-ΔHep-Fc and biotin-HABP with GAGs

A competitive inhibition assay was performed to evaluate the cross-reactivity of biotin-TSG-6-ΔHep-Fc and biotin-HABP with other GAGs (Fig. 3). Binding of biotin-TSG-6-ΔHep-Fc and biotin-HABP to immobilized high molecular weight HA was measured after pre-incubation of the probes with either HA, heparin or chondroitin sulfates A or C. Pre-incubation with HA and, to a lesser extent, with chondroitin sulfates reduced subsequent binding of biotin-TSG-6-ΔHep-Fc and biotin-HABP to immobilized HA, which allowed us to estimate the relative binding avidity by comparing the IC50s for each pair of GAGs (Fig. 3A). When comparing soluble HA and chondroitin sulfate A, there was a 102-fold and a 168-fold difference in IC50s for biotin-TSG-6-ΔHep-Fc and biotin-HABP, respectively. The difference was even greater, in the order of 25,000-fold (biotin-HABP) and over 1000-fold (biotin-TSG-6-ΔHep-Fc), when comparing soluble HA and chondroitin sulfate C. No competitive inhibition was observed when biotin-TSG-6-ΔHep-Fc and biotin-HABP were pre-incubated with up to 4 µM heparin (Fig. 3B). These results demonstrate that the cross-reactivity of biotin-TSG-6-ΔHep-Fc and biotin-HABP towards other GAGs is minimal.

Figure 3.

(A) Competitive inhibition of biotin-HABP (grey squares) or biotin-TSG-6-ΔHep-Fc (black circles) binding to immobilized 1.2 MDa HA after pre-incubation with increasing concentrations of HA, as compared with chondroitin sulfate A (CS A) and chondroitin sulfate C (CS C). Each datapoint is the average of triplicate assays with SEM. Four-parameter inhibition curves and IC50s were calculated using GraphPad Prism. (B) Competitive inhibition of biotin-HABP (grey squares) or biotin-TSG-6-ΔHep-Fc (black circles) binding to immobilized 1.2 MDa HA after pre-incubation with increasing concentrations of heparin. Each datapoint is the average of duplicate assays with SEM.

Size-dependent Competitive Inhibition of HA Binding to Biotin-TSG-6-ΔHep-Fc and Biotin-HABP

To more thoroughly examine the HA binding properties of biotin-TSG-6-ΔHep-Fc and biotin-HABP, their binding to immobilized high molecular weight HA was measured in the presence of polydisperse HA preparations of different molecular weights: 50 kDa (41–65 kDa), 5 kDa (< 10 kDa), oligo-HA₁₀ (5 disaccharide units, 2 kDa), -HA₈ (4 disaccharide units, 1.6 kDa), -HA₆ (3 disaccharide units, 1.2 kDa) and -HA₄ (2 disaccharide units, 0.8 kDa) (Fig. 4). Inhibition of biotin-HABP binding to immobilized HA was dependent on the size of the HA competitor, with no increased competition for HA of 2 kDa (oligo-HA₁₀, 5 disaccharide units) or larger (Fig. 4, left panel). Biotin-TSG-6-ΔHep-Fc binding to immobilized HA was similarly inhibited by HA preparations of 5 kDa or above (Fig. 4, right panel), with oligo-HA₁₀ (2 kDa) being moderately less effective. Oligo-HA₈ and smaller oligo-HA preparations were less efficient as competitors for biotin-TSG-6-ΔHep-Fc or biotin-HABP binding to high molecular weight HA. These results suggest that, with the exception of oligo-HA₁₀, HA binding to biotin-HABP and biotin-TSG-6-ΔHep-Fc are similarly size-dependent.

Figure 4.

Competitive inhibition of high molecular weight HA binding to biotin-HABP (left panel) or biotin-TSG-6-ΔHep-Fc (right panel) after pre-incubation with increasing concentrations of HA preparations of different average molecular weights, as indicated in the legend. Each datapoint is the average of duplicate assays with SEM. Four-parameter inhibition curves were fitted using GraphPad Prism.

Histological Localization of HA using Biotin-HABP and Biotin-TSG-6-ΔHep-Fc in Normal Human Skin

To evaluate whether biotin-TSG-6-ΔHep-Fc could be used for the histolocalization of HA, a procedure similar to the previously described staining with cartilage-derived biotin-HABP (Ripellino et al. 1985; Tammi et al. 1988; Kobayashi et al. 1999) was developed. Biotin-HABP or biotin-TSG-6-ΔHep-Fc probes were detected with peroxidase-streptavidin. Sections of human skin were incubated with either 0.04, 0.2 or 1 µg/ml biotin-HABP or 0.04, 0.2 or 1 µg/ml biotin-TSG-6-ΔHep-Fc (Fig. 5). Pre-incubation with rHuPH20 (Fig. 5, middle panels) or Streptomyces hyalurolyticus hyaluronidase (Fig. 5, right panels) was concurrently performed to assess the specificity of staining and determine optimal concentrations of biotin-HABP or biotin-TSG-6-ΔHep-Fc. At 1 µg/ml, biotin-HABP yielded an intense labeling in the dermis, with some pericellular labeling also present in the epidermis. Pre-incubation of the sections with rHuPH20 or Streptomyces hyalurolyticus hyaluronidase prior to staining resulted in a substantial loss of staining intensity. The residual staining in the dermis likely resulted from the incomplete enzymatic digestion of very high HA concentrations. Residual staining was dependent on the concentration of primary reagent used for HA detection. At a concentration of 0.2 µg/ml, the intensity of the overall staining decreased and the pericellular labeling of the epidermis was lost. Pre-incubation with hyaluronidase almost entirely abolished staining. At a concentration of 0.04 µg/ml, the staining intensity was further decreased and pre-incubation with hyaluronidase resulted in no detectable staining.

Figure 5.

Histological localization of HA in formalin-fixed paraffin-embedded normal human skin using three concentrations of either biotin-HABP (b-HABP) or biotin-TSG-6-ΔHep-Fc (b-TSG-6-ΔHep-Fc) followed with HRP-coupled streptavidin and DAB detection (brown). Gill’s hematoxylin was used as a counterstain (purple). Controls included pre-incubation with recombinant human PH20 (rHuPH20 pretreatment) or Streptomyces hyalurolyticus hyaluronidase (streptomyces hyal pretreatment) and omission of the primary reagents (not shown). Scale Bar = 50 µm.

At 1 µg/ml, biotin-TSG-6-ΔHep-Fc also yielded a very strong labeling in the dermis and in the pericellular spaces of the keratinocytes (Fig. 5). Pre-incubation of the tissue sections with rHuPH20 or Streptomyces hyalurolyticus hyaluronidase prior to staining was insufficient to completely eliminate staining. At a concentration of 0.2 µg/ml, the hyaluronidase-sensitive intense labeling of the dermis, as well as the pericellular labeling of keratinocytes, was comparable to that obtained using 1 µg/ml biotin-HABP. Decreasing the concentration of biotin-TSG-6-ΔHep-Fc further down to 0.04 µg/ml resulted in a reduction in the overall staining intensity and disappearance of epidermal staining similar to what was observed with 0.2 µg/ml biotin-HABP. Omission of biotinylated reagents resulted in no detectable signal (data not shown).

Based on these observations, concentrations of 1 µg/ml biotin-HABP and 0.2 µg/ml biotin-TSG-6-ΔHep-Fc were deemed optimal and selected for further evaluation of the probes and staining methods.

Histolocalization of HA using Biotin-HABP and Biotin-TSG-6-ΔHep-Fc in Human Normal and Malignant Tissues

Biotin-TSG-6-ΔHep-Fc was further characterized as a probe for tissue HA detection and compared with biotin-HABP using TMAs of FFPE normal human tissues (including brain, adrenal, ovary, pancreas, thyroid, parathyroid, pituitary, testis, breast, spleen, tonsil, thymus, bone marrow, lung, cardiac muscle, esophagus, stomach, small intestine, colon, liver, salivary gland, kidney, prostate, skin, uterus, skeletal muscle and nerve) and malignant human tissues from various tumor histotypes. A subset of these tissues is presented in Figure 6. Comparable results were obtained using 1 µg/ml biotin-HABP or 0.2 µg/ml biotin-TSG-6-ΔHep-Fc. High amounts of HA were detected in soft connective tissues including the skin dermis, the stroma of various glands, around blood vessels, in nervous tissues and around muscle fibers. Little to no HA was found in lymphoid tissues, hepatocytes and germinal cells of the testis.

Figure 6.

Histological localization of HA in normal human tissues using biotin-HABP or biotin-TSG-6-ΔHep-Fc. Formalin-fixed paraffin-embedded normal human tissues were stained with 1 µg/ml biotin-HABP (b-HABP) or 0.2 µg/ml biotin-TSG-6-ΔHep-Fc (b-TSG-6-ΔHep-Fc), and visualized with HRP-coupled streptavidin and DAB (brown). Gill’s hematoxylin was used as a counterstain (purple). Controls included pre-incubation with recombinant human PH20 (rHuPH20 pretreatment) and omission of the primary reagents (no primary). Sk. Muscle, skeletal muscle. Scale Bar = 200 µm.

Pre-incubation of the tissue sections with rHuPH20 prior to staining was used as an analytical specificity control and resulted in substantially reduced HA staining intensity or a complete loss, depending on initial staining intensity. Omission of biotinylated reagents resulted in no detectable signal.

Comparison of Biotin-HABP and Biotin-TSG-6-ΔHep-Fc Staining Intensity using Digital Image Analysis

In order to objectively compare biotin-HABP and biotin-TSG-6-ΔHep-Fc staining intensity in normal and malignant human tissues, slides were scanned and staining positivity was digitally measured using the Aperio Positive Pixel Count v9 algorithm, as described in the Materials & Methods. A very strong linear correlation was found between biotin-HABP and biotin-TSG-6-ΔHep-Fc staining positivity, with a Pearson coefficient of 0.96 for 125 pairs (Fig. 7), indicating that the two reagents are comparable with respect to their diagnostic specificity for HA detection in human tissues.

Figure 7.

Correlation analysis between biotin-HABP (b-HABP) and biotin-TSG-6-ΔHep-Fc (b-TSG-6-ΔHep-Fc) staining positivity. Positivity is defined by the ratio of pixels above a defined intensity threshold relative to the total pixels stained and counterstained as measured using the Aperio Positive Pixel Count algorithm for each sample. Each datapoint represents one of 125 pairs of serial normal or malignant tissue core sections.

Discussion

Biotinylated HABP preparations derived from animal nasal cartilage have been widely used for the detection of HA in various types of samples, including cells and tissue sections (Laurent et al. 1991). However, these commercial preparations are sub-optimal for the development of a robust and reliable semi-quantitative HA-detection method. First, the binding of cartilage-derived biotin-HABP (which consists of a mixture of proteins) to HA is a complex process, and involves both protein-HA and protein-protein (aggrecan-cartilage link protein) interactions. Moreover, because of its multi-component nature, it is not possible to obtain quantitative and accurate binding constants. Finally, due to the low purity and potential batch-to-batch variability, cartilage-derived HABP is not suitable for precise clinical applications such as cancer diagnostics. On the other hand, recently developed HA-specific probes, such as the neurocan HA-binding domain fused with green fluorescent protein (GFP) (Zhang et al. 2004), the PEP-1 synthetic peptide (Zmolik and Mummert 2005) and the recombinant versican G1 domain (Clark et al. 2011) await more thorough characterization.

To address these limitations, a stable, reliable and specific probe consisting of a recombinant fusion protein, produced in mammalian cells and composed of a modified TSG-6 Link module fused to the Fc portion of human IgG1 was developed for the in situ detection of HA in tissues. The TSG-6 Link module was selected for its relatively high HA binding affinity (Kahmann et al. 2000; Lesley et al. 2002) and the addition of the human constant (Fc) domain of immunoglobulin G1 (IgG1) to the TSG-6 Link module allowed for simple purification and detection steps. Because previous studies have shown that the Link module of TSG-6 also contains a heparin-binding domain (Mahoney et al. 2005), a low heparin-binding TSG-6-Fc (TSG-6-ΔHep-Fc) was obtained by substituting three lysine residues for alanine residues (K20A/K34A/K41A) in the heparin-binding domain as previously described (Mahoney et al. 2005) in order to enhance specificity for HA while maintaining a HA binding affinity similar to that of the parental molecule (TSG-6-Fc); this is because HA and heparin-binding sites are located in different regions of the Link module (Mahoney et al. 2001; Mahoney et al. 2005). The heparin-binding activity of TSG-6-ΔHep-Fc was decreased by 90% as compared with TSG-6-Fc, consistent with previous results obtained with TSG-6 Link module monomers harboring similar mutations (Mahoney et al. 2005). This substitution therefore minimizes any potential cross-reactivity with heparin/heparan sulfate present in tissues and thus increases its specificity for HA.

TSG-6-ΔHep-Fc was coupled to biotin to allow histological detection with standard fluorescent or colorimetric secondary streptavidin reagents for visualization of HA in the cell matrix. Although HA staining intensity in tissue sections can vary depending on the fixation procedure (Lin et al. 1997), clinical specimens, such as tumor biopsies, are routinely fixed in formalin at the clinical site, to be processed and embedded into paraffin for histological examination. For this reason, our evaluation of biotin-TSG-6-ΔHep-Fc was performed on FFPE human tissues only, although TSG-6-ΔHep-Fc has also been successfully used to label HA in cells in culture by immunofluorescence (data not shown).

Biotin-HABP and biotin-TSG-6-ΔHep-Fc both displayed little cross-reactivity towards chondroitin sulfate A, another GAG which is structurally similar to HA, with at least 100-fold more avidity towards HA than chondroitin sulfate A, and minimal cross-reactivity towards chondroitin sulfate C and heparin

The minimum molecular size of HA oligosaccharides that were bound by biotin-HABP and biotin-TSG-6-ΔHep-Fc was 2 kDa (5 disaccharide units), as assessed by our competitive binding assay to immobilized high molecular weight HA. There was no molecular weight-dependent increase in binding efficiency of either biotin-HABP or biotin-TSG-6-ΔHep-Fc to HA of 5 kDa and above. Binding to HA of 5 disaccharide units (2 kDa) was less efficient for biotin-TSG-6-ΔHep-Fc than for biotin-HABP as compared with their respective binding to higher molecular weight HA. The strong linear correlation between biotin-HABP and biotin-TSG-6-ΔHep-Fc staining intensity in normal and malignant human tissues suggests that this does not significantly affect HA detection in tissues. Neither biotin-HABP nor biotin-TSG-6-ΔHep-Fc was found to bind efficiently to HA of less than 2 kDa.

The histological detection of HA in FFPE tissue sections using biotin-TSG-6-ΔHep-Fc resulted in hyaluronidase-sensitive staining patterns that were comparable with those obtained with biotin-HABP, but with greater sensitivity, since approximately 5-fold less biotin-TSG-6-ΔHep-Fc than biotin-HABP was necessary to achieve a similar staining intensity. Analytical specificity was assessed by pre-incubating tissues with hyaluronidase in order to remove HA prior to staining. To demonstrate specificity of staining, sections were treated with Streptomyces hyalurolyticus hyaluronidase, a very specific HA-degrading enzyme with low activity towards chondroitin sulfate (Yamada 1973; Derby and Pintar 1978), as well as with rHuPH20, an enzyme belonging to the mammalian family of hyaluronidases, which have been shown to degrade chondroitin sulfate to a certain extent (Kreil 1995, Rappaport et al. 1951; Frost 2007). Predigestion of tissue HA with rHuPH20 yielded results similar to Streptomyces hyalurolyticus hyaluronidase in both normal and malignant human tissues, but required shorter incubation times. Hyaluronidase treatment resulted in a complete loss of staining with biotin-HABP or biotin-TSG-6-ΔHep-Fc probes in most tissues, except in tissues where HA content was very high prior to digestion, such as in the skin and in the stroma of several glands (Wang et al. 1996; Thompson et al. 2010; Jiang et al. 2012). These tissues exhibited very strong HA labeling intensity when stained with biotin-HABP or biotin-TSG-6-ΔHep-Fc.

Taken together, these results demonstrate that biotin-TSG-6-ΔHep-Fc is more sensitive than and as specific as biotin-HABP for HA detection in human tissues. Specific, intense HA staining can be achieved with as little as 0.2 µg/ml biotin-TSG-6-ΔHep-Fc when combined with direct peroxidase/streptavidin detection, which is less than the concentration needed for most antibodies to achieve satisfactory antigen detection. This HA probe has been used to perform dual fluorescent labeling in conjunction with an anti-CD44 antibody (data not shown). In summary, the present work describes the generation and characterization of a sensitive and specific HA-binding probe and demonstrates its value for histological detection of HA in FFPE human tissues. This is an essential step towards developing a sensitive and specific companion diagnostic device to characterize HA in individual tumors as a means of selecting patients who are more likely to respond to HA-depleting treatments. The ability to specifically label and image HA in malignant tissues will also contribute to understanding the role of HA in normal physiology and development as well as in disease.

Footnotes

Acknowledgements

The authors wish to thank Shiping Fang for the LC-MS analysis of TSG-6-ΔHep-Fc.

Declaration of Conflicting Interests

All authors are or were employees of Halozyme Therapeutics at the time the work was performed.

Funding

The authors and Halozyme Therapeutics received no external financial support for the research, authorship, and/or publication of this article.

References

Baier

Baader

Jankowski

Gieselmann

Schilling

Rauch

Kappler

(2007). Hyaluronan is organized into fiber-like structures along migratory pathways in the developing mouse cerebellum. Matrix Biol 26:348-358.

Baker

Bookbinder

(2011). Large-scale production of hyaluronidase related applications. Halozyme, Inc. US Patent 20,110,053,247, Publication number EP2268805 A1.

Clark

Keenan

Fielder

Collinson

Holley

Merry

van Kuppevelt

Day

Bishop

(2011). Mapping the differential distribution of glycosaminoglycans in the adult human retina, choroid, and sclera. Invest Ophthalmol Vis Sci 52:6511-6521.

Derby

Pintar

(1978). The histochemical specificity of Streptomyces hyaluronidase and chondroitinase ABC. Histochem J 10:529-547.

Evanko

SP and

Wight

(1999). Intracellular localization of hyaluronan in proliferating cells. J Histochem Cytochem 47:1331-1342.

Frost

(2007). Recombinant human hyaluronidase (rHuPH20): an enabling platform for subcutaneous drug and fluid administration. Expert Opin Drug Deliv 4:427-440.

Fülöp

Szántó

Mukhopadhyay

Bárdos

Kamath

Rugg

Day

Salustri

Hascall

Glant

Mikecz

(2003). Impaired cumulus mucification and female sterility in tumor necrosis factor-induced protein-6 deficient mice. Development 130:2253-2261.

Giudicelli

Chaume

Lefranc

(2005). IMGT/GENE-DB: A comprehensive database for human and mouse immunoglobulin and T cell receptor genes. Nucleic Acids Res 33(Database issue):D256-D261.

Hascall

Majors

De La Motte

Evanko

Wang

Drazba

Strong

(2004). Intracellular hyaluronan: a new frontier for inflammation? Biochim Biophys Acta 1673:3-12.

10.

Itano

Kimata

(2002). Mammalian hyaluronan synthases. IUBMB Life 54:195-199.

11.

Itano

Zhuo

Kimata

(2008). Impact of the hyaluronan-rich tumor microenvironment on cancer initiation and progression. Cancer Sci 99:1720-1725.

12.

Jacobetz

Chan

Neesse

Bapiro

Cook

Frese

Feig

Nakagawa

Caldwell

Zecchini

Lolkema

Jiang

Kultti

Thompson

Maneval

Jodrell

Frost

Shepard

Skepper

Tuveson

(2012). Hyaluronan impairs vascular function and drug delivery in a mouse model of pancreatic cancer. Gut 62:112-120.

13.

Jadin

Bookbinder

Frost

(2012). A comprehensive model of hyaluronan turnover in the mouse. Matrix Biol 31:81-89.

14.

Jiang

Thompson

Huang

Araiza

Osgood

Wei

Feldmann

Frost

Shepard

(2012). Effective targeting of the tumor microenvironment for cancer therapy. Anticancer Res 3:1203-1212.

15.

Kahmann

O’Brien

RRO

Werner

HeinegardHeinegård

Ladbury

Campbell

Day

(2000). Localization and characterization of the hyaluronan-binding site on the link module from human TSG-6. Structure 8:763-774.

16.

Kobayashi

Sun

Terao

(1999). Immunolocalization of hyaluronic acid and inter-alpha-trypsin inhibitor in mice. Cell Tissue Res 296:587-597.

17.

Kohda

Morton

Parkar

Hatanaka

Inagaki

Campbell

Day

(1996). Solution structure of the link module: A hyaluronan-binding domain involved in extracellular matrix stability and cell migration. Cell 86:767-775.

18.

Kreil

(1995). Hyaluronidases – A group of neglected enzymes. Protein Sci 4:1666-1669.

19.

Kuznetsova

Mahoney

Martin-Manso

Ali

Nentwich

Sipes

Zeng

Vogel

Day

Roberts

(2008). TSG-6 binds via its CUB_C domain to the cell-binding domain of fibronectin and increases fibronectin matrix assembly. Matrix Biol 27:201-210.

20.

Kultti

Jiang

Thompson

Frost

Shepard

(2012). Therapeutic Targeting of Hyaluronan in the Tumor Stroma. Cancers (Basel) 4:873-903.

21.

Laurent

Johnson-Well

Hellström

Engström-Lauren

Well

(1991). Localization of hyaluronan in various muscular tissues. A morphological study in the rat. Cell Tissue Res 263:201-205.

22.

Lee

Wisniewski

Vilcek

(1992). A novel secretory tumor necrosis factor-inducible protein (TSG-6) is a member of the family of hyaluronate binding proteins, closely related to the adhesion receptor CD44. J Cell Biol 116:545-557.

23.

Lesley

English

Gal

Mikecz

Day

Hyman

(2002). Hyaluronan-binding properties of a CD44 chimera containing the link module of TSG-6. J Biol Chem 277:26600-26608.

24.

Lin

Shuster

Maibach

Stern

(1997). Patterns of hyaluronan staining are modified by fixation techniques. J Histochem Cytochem 45:1157-1163.

25.

Lindqvist

Chichibu

Delpech

Goldberg

Knudson

Poole

Laurent

(1992). Seven different assays of hyaluronan compared for clinical utility. Clin Chem 38:127-132.

26.

Mahoney

Blundell

Day

(2001). Mapping the hyaluronan-binding site on the link module from human tumor necrosis factor-stimulated gene-6 by site-directed mutagenesis. J Biol Chem 276:22764-22771.

27.

Mahoney

Mulloy

Forster

Blundell

Fries

Milner

Day

(2005). Characterization of the interaction between tumor necrosis factor-stimulated Gene-6 and heparin: Implications for the inhibition of plasmin in extracellular matrix microenvironments. J Biol Chem 280:27044-27055.

28.

Milner

Day

(2003). TSG-6: a multifunctional protein associated with inflammation. J Cell Sci 116:1863-1873.

29.

Milner

Higman

Day

(2006). TSG-6: a pluripotent inflammatory mediator? Biochem Soc Trans 34:446-450.

30.

Provenzano

Cuevas

Chang

Goel

Von Hoff

Hingorani

(2012). Enzymatic targeting of the stroma ablates physical barriers to treatment of pancreatic ductal adenocarcinoma. Cancer Cell 21:418-429.

31.

Rappaport

Myer

Linker

(1951). Analysis of the products formed on hydrolysis of hyaluronic acid by testicular hyaluronidase. J Am Chem Soc 73:2416-2420.

32.

Ripellino

Klinger

Margolis

(1985). The hyaluronic acid binding region as a specific probe for the localization of hyaluronic acid in tissue sections. J Histochem Cytochem 33:1060-1066.

33.

Sakai

Yasuda

Sayo

Ishikawa

Inoue

(2000). Hyaluronan exists in the normal stratum corneum. J Invest Dermatol 114:1184-1187.

34.

Selbi

Day

Rugg

Fülöp

De La Motte

Bowen

Hascall

Phillips

(2006). Overexpression of hyaluronan synthase 2 alters hyaluronan distribution and function in proximal tubular epithelial cells. J Am Soc Nephrol 17:1553-1567.

35.

Simpson

Meran

Thomas

Stephens

Bowen

Steadman

Phillips

(2009). Age-related changes in pericellular hyaluronan organization leads to impaired dermal fibroblast to myofibroblast differentiation. Am J Pathol 175:1915-1928.

36.

Stern

Asari

Sugahara

(2005). Hyaluronan fragments: an information-rich system. Eur J Cell Biol 85:699-715.

37.

Tammi

Ripellino

Margolis

Tammi

(1988). Localization of epidermal hyaluronic acid using the hyaluronate binding region of cartilage proteoglycan as a specific probe. J Invest Dermatol 90:412-414.

38.

Tengblad

(1979). Affinity chromatography on immobilized hyaluronate and its application to the isolation of hyaluronate binding properties from cartilage. Biochim Biophys Acta 578:281-289.

39.

Thompson

Shepard

O’Connor

Kadhim

Jiang

Osgood

Bookbinder

Sugarman

Connor

Nadjsombati

Frost

(2010). Enzymatic depletion of tumor hyaluronan induces antitumor responses in preclinical animal models. Mol Cancer Ther 9:3052-3064.

40.

Toole

(2004). Hyaluronan: from extracellular glue to pericellular cue. Nat Rev Cancer 4:528-539.

41.

Toole

(2009). Hyaluronan-CD44 Interactions in Cancer: Paradoxes and Possibilities. Clin Cancer Res 15:7462-7468.

42.

Wang

Tammi

Guo

Tammi

(1996). Hyaluronan distribution in the normal epithelium of esophagus, stomach, and colon and their cancers. Am J Pathol 148:1861-1869.

43.

Weigel

Hascall

Tammi

(1997). Hyaluronan synthases. J Biol Chem 272:13997-14000.

44.

Wells

Lundin

Michaelsson

(1991). Histochemical localization of hyaluronan in psoriasis, allergic contact dermatitis and normal skin. Acta Derm Venereol 71:232-238.

45.

Yamada

(1973). The effect of digestion with Streptomyces hyaluronidase upon certain histochemical reactions of hyaluronic acid-containing tissues. J Histochem Cytochem 21:794-803.

46.

Zhang

Baader

Sixt

Kappler

Rauch

(2004). Neurocan-GFP fusion protein: a new approach to detect hyaluronan on tissue sections and living cells. J Histochem Cytochem 52:915-922.

47.

Zmolik

Mummert

(2005). Pep-1 as a Novel Probe for the In Situ Detection of Hyaluronan. J Histochem Cytochem 53:745-751.