Selective Isolation of Hyaluronan by Solid Phase Adsorption to Silica

Abstract

A solid phase adsorption method for selective isolation of hyaluronan (HA) from biological samples is presented. Following enzymatic degradation of protein, HA can be separated from sulfated glycosaminoglycans, other unsulfated glycosaminoglycans, nucleic acids, and proteolytic fragments by adsorption to amorphous silica at specific salt concentrations. The adsorbed HA can be released from silica using neutral and basic aqueous solutions. HA ranging in size from ~ 9 kDa to MDa polymers has been purified by this method from human serum and conditioned medium of cultured cells.

Graphical Abstract

1. Introduction

Hyaluronan (HA), also called hyaluronic acid, is a linear glycosaminoglycan polymer composed solely of repeating disaccharides of [(1→4)-β-D-GlcA-(1→3)-β-D-GlcNAc]. It is a major component of the extracellular matrix in many vertebrate tissues, where its physiological functions are closely linked with its molecular mass (M). Determination of M distribution for HA, and its changes with pathological processes, is of high interest. The normal M distribution of naturally occurring HA is polydisperse, averaging approximately 2-6 million Daltons (2000-6000 kDa) in many healthy tissues, but ranging from perhaps 10,000 kDa polymers to oligosaccharide fragments with M of a few kDa [1-7].

HA plays a major role in homeostasis of the cellular microenvironment. Long chains of HA are synthesized by three transmembrane hyaluronan synthase isozymes: HAS1, HAS2 and HAS3, and the rate of HA synthesis appears to be increased under many inflammatory conditions[8-10]. In healthy tissues, long HA chains are tethered to clustered cell surface CD44 receptors and other hyaladherins (HA-binding proteins), and serve, in turn, to bind and organize proteoglycans of the pericellular matrix[11-13]. The clustered HA-CD44 motif controls CD44-mediated cytoplasmic signaling pathways, and dampens inflammatory response to interleukin-1β (IL-1β), tumor necrosis factor-alpha (TNF-α), and lipopolysaccharide (LPS)[14-16]. High M HA also protects the cell membrane by reacting with free radicals and other reactive oxygen and nitrogen species, at the cost of chain breaks that reduce the HA size and CD44 clustering[17-20]. HA fragmentation can also result from the action of secreted or shed hyaluronidases[21, 22]. In humans, HA is primarily degraded by the actions of TMEM2, CEMIP (also called HYBID or KIAA1199), Hyal2, and Hyal1[21-26]. The overall balance of synthesis and degradation controls the content and size of HA in tissues and fluids of the body[7].

The degradation of HA is of particular interest in cell signaling. HA fragments can act as reporters of damaging processes in the cellular microenvironment, and have been found to stimulate pro-inflammatory or defensive responses in some cells and tissues[27-33]. The receptors sensing the presence of low M HA are known to include CD44, which can become de-clustered, RHAMM, which binds CD44 and low M HA simultaneously, and Toll-like receptors 2 and 4, which may detect HA fragments indirectly by their effect on CD44[20]. Signaling by those receptors in response to HA fragments leads to diverse cellular responses associated with cell migration, tissue remodeling and repair, development, tumor growth and metastasis, response to microbial infection, and many other processes[34, 35]. In other cells and tissues, HA fragments do not appear to be pro-inflammatory; this apparent discrepancy may reflect tissue-specific and context-specific differences in levels or activation status of receptors and signal pathway proteins[36-38].

The M distribution of HA is important in both normal physiological and pathological processes, therefore researchers need to be able to isolate and analyze all sizes of HA molecules from a wide variety of biological fluids and tissues, as well as many types of cultured cells used to model the processes. The object is to isolate HA quantitatively and without degradation during processing, preserving the full M distribution present in a given tissue at a given time.

Multiple methods to isolate HA have been developed. HA isolation was first achieved by Karl Meyer in 1934 [39] from bovine vitreous humor, and in 1936 [40] from human umbilical cords. In 1956, Meyer and coworkers reviewed the glycosaminoglycan (GAG) composition of numerous tissues, and described optimized methods for extraction[41]. GAG and HA extraction generally included the following steps: dehydration of the tissue source with acetone to remove contaminants while keeping GAGs insoluble, then choosing to extract using either (i) 0.33 M NaOH (to liberate sulfated GAGs from core proteins), (ii) water, (iii) 2% aqueous phenol, and/or (iv) proteolytic digestion with pepsin at low pH, then trypsin at neutral pH. HA was separated from sulfated GAGs by differential precipitation from calcium acetate solutions with ethanol, or by anion exchange chromatography.

In the next few decades, other researchers expanded on these methods to isolate HA from different biological fluids and tissues such as synovial fluid, plasma, urine, brain, and muscle [42-48]. Tissues and fluids were often frozen or freeze-dried before being used. Prior to extracting HA, some tissues or biological fluids were treated to reduce lipid content while inactivating hydrolytic enzymes. Chloroform and methanol or ethanol have been used, but also acetone and ether in some cases. Protein was usually removed by proteolytic digestion, or alternatively by precipitation with perchloric acid, phosphotungstic acid, or trichloroacetic acid. Finally, HA was usually separated from sulfated GAGs by (i) fractional precipitation with ethanol, (ii) by complexation with cationic quaternary ammonium species like cetyl pyridinium chloride, and/or (iii) by anion exchange chromatography.

Isolation of high molecular weight HA with sufficiently high purity for medical use was described in a 1979 patent by Balazs [49], using HA-rich rooster comb as source material. Blood products were removed from the tissue using chloroform-denatured ethanol, then HA was extracted with water containing 5% chloroform. Protein removal employed several steps. First, NaCl was added to 1 M or 10% to dissociate proteins from HA, then the solution was filtered through sintered glass or nylon, and an equal volume of chloroform was added to denature the protein and collect it at the aqueous-organic interface. To further remove protein, the aqueous layer was acidified to pH 4.5 with 1 M HCl, and re-extracted with chloroform. Pronase and DNase could also be used to enzymatically remove protein and DNA. Removal of enzymes was performed after neutralization, by re-extraction with chloroform for 7 days. The extracted HA solution was sterile filtered, then HA precipitated with acetone or ethanol, and redissolved in 0.1 M NaCl. This process was repeated as needed.

Similar time-consuming methods are still employed for bioanalytical applications requiring determination of the molecular weight distribution or content of HA [6, 9, 30, 38, 50-61]. The isolation steps usually employ organic solvents for protein denaturation and/or HA precipitation, multiple enzymatic degradations, solvent changes and removal of small solute contaminants by dialysis, concentration, additional precipitations, and re-dissolution steps. The time needed for most accepted protocols is multiple days and often require large sample amounts (>10-100 milligrams) due to the multitude of processing steps. Even then, the HA can be accompanied by impurities, mainly enzyme-resistant proteins but also the relatively rare unsulfated glycosaminoglycan chondroitin[30, 56]. Specific isolation of HA using streptavidin-labeled beads and a biotinylated HA-binding protein has also been developed, but the yield is very low[56, 62].

If the HA is radiolabeled or will be detected by its binding to specific detector proteins, it may not need to be exhaustively purified, but it must still be liberated from endogenous protein binding partners that can interfere with HA separation or detection.

As a further concern, most of the procedures currently employed have never been validated for quantitative yield, and perhaps more importantly, for faithful maintenance of the original size distribution of the HA present in the sample. HA is extremely susceptible to degradation by redox reactions catalyzed by metal ions (which can be present in solvents stored in metal containers, and can also exist in tissue extracts) and by reagents such as cysteine used to activate certain proteases (e.g., papain) [20].

Based on these obstacles, we determined that there was a need for a faster and more simple method, in which yield could be high and size distribution retained. Just as facile isolation of DNA and RNA revolutionized molecular biology, a similarly simple method for isolation of HA could have a significant impact on scientific research leading to an improved understanding of the role of HA in cellular response to changes in the surrounding microenvironment of the extracellular matrix.

For DNA and RNA isolation, amorphous silica particles or membranes and glass fibers have been used as adsorbants since the discovery that DNA is adsorbed on glass in concentrated NaI solutions used to disrupt agarose gels after electrophoresis[63]. Further studies have shown that DNA and RNA can be adsorbed to silica in weakly acidic solutions containing chaotropic agents such as NaClO₄, GuSCN, and GuHCl, sometimes with addition of a nonsolvent like ethanol to increase adsorption[64-67], and these solutes are commonly used in nucleic acid isolation kits. Considering that HA is also a polyanion, we attempted to isolate HA using multiple commercially available kits for DNA and RNA isolation. We found that HA did not bind silica under those nucleic acid-optimized conditions. We therefore undertook to discover conditions under which HA could be selectively bound to a simple substrate like silica and simultaneously separated from other types of glycosaminoglycans, nucleic acids, and proteins or proteolytic fragments.

2. Materials and Methods

2.1. Chemicals

Silica was obtained as amorphous precipitated silica gel particles, 40-63 μm diameter, 60 Å pore size, surface area averaging 484 m²/g (Silicycle # R10030B). We have also used Sigma silica (#S5631) fine particles derived from naturally occurring microcrystalline silica ground to a powder, with diameter 0.5-10 μm (80% with diameter 1-5 μm) (may be used in bulk separation by centrifugation but not suitable for use on a spin column with a 20 μm average pore size filter). HA samples were HA1700 kDa (Lifecore Biomedical, LLC, Chaska MN), SelectHA™ 31kDa, 50kDa, 150kDa (Hyalose LLC, Oklahoma City, OK, and Echelon Biosciences, Salt Lake City, UT), and HA 9kDa, a low molecular mass HA sample containing chains 19–25 disaccharides in length (8–10 kDa) prepared as previously described[68]. Sulfated glycosaminoglycans were chondroitin sulfate (Calbiochem # 230699)(average M approximately 40 kDa), heparin (Sigma #H4784)(average M approximately 16 kDa), and heparan sulfate (Sigma #H7640)(average M approximately 40 kDa). The average M for each sulfated GAG sample was estimated by co-electrophoresis with known M standards (Supplementary Figure S1). Unsulfated glycosaminoglycans chondroitin (polydisperse, mainly 100-200 kDa, from bacterial fermentation) and heparosan (quasi-monodisperse, approximately 100 kDa, from chemoenzymatic synthesis) were prepared as in[69] and[70] , respectively. Nucleic acid samples were DNA restriction fragments, 100 bp ladder 100-1517 bp (New England Biolabs #N3231S) and ssRNA ladder containing RNA of 500-9000 bases (New England Biolabs #0362S). Protease was Proteinase K (Sigma #3115887001, 18.5 mg/ml in 10 mM Tris pH 7.5, stabilized with calcium acetate). Ethanol, 200 Proof, ACS Certified, was from Sigma Aldrich. Trifluoroethanol was from Chem-Impex Intl., # 04832. Phosphate-buffered saline (PBS, 0.01 M phosphate, 0.138 M NaCl, 0.0027 M KCl, pH 7.4) was from Sigma Chemical, #P3813, and was filtered using a Corning 430769, 0.22 μm CA sterilizing filter system. All other chemicals were reagent grade. Deionized water, 0.2 μm filtered, was from a MilliQ water purification system.

2.2. Biological fluids

Conditioned medium was from HEK293 cells (ATCC), grown and passaged in DMEM containing 10% FBS and 1% penicillin-streptomycin. Human type AB serum (sterile and HCV, HIV, HBsAG free) was from MP Biomedical, #092931949.

2.3. Devices and kits

Mini Spin Columns, 0.7 mL bed volume with a 20 μm pore polyethylene filter, were from BioVision (# 6572). Zeba Spin desalting columns, 2 mL, 7K MWCO, were from Life Technologies (# 89890). Aurum Affi-Gel Blue mini columns were from Bio-Rad Laboratories (#7326708). DNA and RNA isolation kits were PureLink Viral RNA/DNA mini kit (Thermo Fisher #12280050), MagMax 96 DNA Multi-Sample Kit (Thermo Fisher #4413021), HighPure PCR Template Preparation Kit (Roche #11796828001), MagnaPure LC DNA Isolation Kit (Roche Diagnostics #03003990001), RNeasy kit (Qiagen #74004). High capacity endotoxin removal columns were from ThermoFisher Pierce (#88274). Poly-L-lysine-Separopore 4B-CL (polylysine-agarose) #20181075-3 was from Bioworld (Dublin, OH). Polyacrylamide gels for electrophoresis were Novex 4-20% polyacrylamide gradient 12-well gels from Life Technologies (#EC62252BOX). Other small equipment used included Thermo Scientific Sorvall Legend Micro 21 microcentrifuge; SCILOGEX SCI-T6-S, tube roller (½ max speed=40rpm); Thermo Scientific Savant DNA 130 Vacuum Concentrator System (Cat. # 13 875 331).

2.4. Isolation Procedure

2.4.1. Pretreatment of samples

HA-containing samples are processed to remove the majority of proteins before the silica isolation step.

2.4.1.1. pretreatment of conditioned medium

To provide visual validation of HA isolation without degradation, a duplicate sample of the conditioned medium sample may be spiked with a mixture of GAGs. Per 240 μL of conditioned medium, add (all control GAG solutions were 1 μg/μL in water) 1μg HA 31kDa + 1μg HA 50kDa + 1μg HA 150kDa + 1μg HA 1.7MDa+ 2μg Chondroitin Sulfate + 2μg Heparan Sulfate + 1μg Heparin + 1μL filtered diH₂O = 10μL. To digest proteins in a 250 μL sample, 10 μL proteinase K solution containing 185 μg enzyme is added and the sample incubated at 60°C for 3 h. It is then boiled for 15 min, cooled, and clarified by centrifugation through an empty spin column at 1500 x g for 5 min. The sample is weighed to determine volume, and dry NaCl added to 5 M final concentration.

2.4.1.2. pretreatment of serum

To provide visual validation of HA isolation without degradation, a duplicate sample of serum may be spiked with HA. Per 100 μL of serum, add (all control HA solutions were 1 μg/μL in water) 1μg HA 9kDa + 0.5μg HA 50kDa + 0.5μg HA 150kDa + 0.5 μg HA 1.7MDa + 7.5μL filtered diH₂O = 10μl, and then mixed with 300 μL of 20 mM sodium phosphate buffer pH 7.0 (=dilute phosphate buffer). Aurum Affi-Gel Blue mini spin column prewashed with dilute phosphate buffer was prepared. The spin columns contain 0.45 ml of crosslinked agarose with covalently attached Cibacron Blue F3GA dye. Each column can bind up to about 5 mg albumin (mammalian sera contains ~35-50 mg/ml albumin). Load the 400 μL of diluted serum on a column, allow to drain by gravity into a clean 1.5 mL microcentrifuge tube for 15 min, then collect the unbound solution by centrifugation at 10,000 x g for 20 s, and repeat with an additional 400 μL of dilute phosphate buffer to wash the column without waiting for a gravity flow elution step. Combine the two unbound solutions. It is possible to isolate HA directly from this sample by adding dry NaCl to make the solution 5M, then adsorbing to silica for the isolation steps described below. Because the remaining protein content can partially block HA access to the silica surface, it is preferable to digest protein first. The sample is solvent exchanged to 0.1 M NaCl using a ZebaSpin column pre-equilibrated in that solvent, then treated with 40 uL proteinase K for the 800 uL sample, at 60°C for 60 min. It is then boiled for 15 min, cooled, and clarified by centrifugation through an empty spin column at 1500 x g for 5 min. The sample is weighed to determine volume, and dry NaCl added to 5 M final concentration.

2.4.2. HA adsorption and elution from silica

Pretreated samples of Step 1 are bound, washed, then eluted for HA isolation. All reagents were used at room temperature, unless noted otherwise.

2.4.2.1. Silica Preparation

Prepare a 10% amorphous silica suspension in water, washing twice with centrifugation at 400 x g for 5 min to remove fine particles. Store at room temperature, adding 0.05% sodium azide if longer than two weeks. Re-suspend by brief vortexing and place on a roller for 10 min prior to use. Prepare an empty mini spin column (e.g., with 700 μL bed volume and 20 μm pore size polyethylene support). Twist off (and save) the fixed plug of the spin column. Remove the screw cap top. Place the spin column into a 1.5-2.0 mL centrifuge tube. Transfer 700 μL of the 10% silica to the column. Centrifuge at 1500 x g for 5 min. Discard the flow through water. Add an additional 700 μL of 10% silica and repeat. Replace the bottom plug on the spin column. Add 240 μL 5M NaCl and replace screw cap top. Vortex for 10 seconds to mix to pre-equilibrate the silica. Remove (and save) bottom plug and place the mini spin column back into a centrifuge tube. Loosen the screw cap and centrifuge at 1500 x g for 5 minutes. Replace bottom plug.

2.4.2.2. Sample Loading

Add ~240 μL sample in 5 M NaCl, tighten cap and vortex 10 s. Incubate at room temperature for 20 min on horizontal roller. Remove bottom plug, place spin column into a clean 1.5 mL centrifuge tube with the top screw cap loosened. Centrifuge at 1500 x g for 5 min to collect the first unbound fraction. Replace bottom plug on the mini spin column. Add 240 μL 5 M NaCl and screw cap on. Vortex for 10 s to mix. Incubate on roller for 20 min, and centrifuge to collect second unbound fraction. Save unbound samples if isolation of sulfated GAGs is desired.

2.4.2.3. Wash – removal of unbound components

Wash the silica twice with 240 μL of 50% trifluoroethanol in 5 mM Tris, 0.5 mM EDTA, pH 7.1, following the same steps of section 2.4.2.2, including brief vortex, 20 min incubation with rolling, and centrifugation. CAUTION: stocks of trifluoroethanol are moderately toxic and should be handled with gloves and manipulated in a chemical hood. Replacement of 50% trifluoroethanol with 75% ethanol in the same Tris-EDTA buffer can be used, but there is some loss of the lowest M HA (9-50 kDa).

2.4.2.4. HA Elution

Following the same mixing, incubation, and centrifugation protocols as in sections 2.4.2.2 and 2.4.2.3, elute the HA from the silica using three sequential steps with three different buffers. Elution buffer 1: 240 μL of 10 mM Tris, 1 mM EDTA, pH 7.5. This step is preferably done twice. Elution buffer 2: 240 μL of PBS, pH 7.4. Elution buffer 3: 240 μL of 0.4 M KOH (This step may be most efficient with an overnight incubation at 4°C, but a 20 min room temperature incubation elutes the majority of the HA). Neutralize the KOH extract using 96 μL 1 M CH₃COOH (preferred method due to buffering capacity) or 96 μL 1M HCl. Pool elutions and desalt using Zeba Spin desalting columns (this may require two columns to accommodate the volume). If desired, dry samples on a centrifugal evaporator using heat up to 60°C. Samples may be re-dissolved in filtered deionized water.

2.4.3. Additional experimental notes

Desalting HA is most easily and rapidly accomplished using spin desalting columns with a 7K MWCO (molecular weight cutoff). We found that desalting or solvent exchange using centrifugal filters of 3-10K MWCO can result in sample loss, depending on the initial solvent. HA in water is sufficiently expanded in hydrodynamic volume to be retained on the filters, but HA in moderate to high ionic strength salt solutions or in buffers containing ethanol or trifluoroethanol is more contracted in molecular volume and thus passes through the semi-porous membranes.

Biological fluids or extracts should generally be pretreated with a protease such as proteinase K. After proteinase K treatment, boiling to denature the enzyme (heating does not cause HA degradation) and either centrifugation at 20,000 x g or filtering through an empty spin column with 20 μm pore size may be used to clarify the sample before proceeding to adding NaCl prior to silica adsorption. The proteinase K digestion solvent should contain an approximately physiological concentration of NaCl.

2.5. Analytical methods

HA quantification was performed using an HA-specific AlphaScreen assay[71]. Gel electrophoresis was performed on 4-20% polyacrylamide gradient gels in Tris-borate-EDTA (TBE) buffer by the method of Bhilocha et al[72], using samples containing ~1-10 μg GAG in 10 μL water, mixed with 2 μL 0.03% bromophenol blue, 2M sucrose in TBE. Gels were electrophoresed at 400V for 22 min, stained with 0.005% Stains-All in 50% ethanol for 90min, and de-stained in 10% ethanol for 10 min. Gels were scanned in grayscale using a red filter on a calibrated ImageScanner III with LabScan v.6, and color images were obtained using Epson scan software. Grayscale image files (16 bit) were densitometrically analyzed using ImageQuant TL 8.2 software (Cytiva).

3. Results

3.1. HA can be reversibly adsorbed to silica in concentrated NaCl solutions

Pure HA, ranging in size from 9 to 1700 kDa, was tested for binding to amorphous precipitated silica gel particles (natural microcrystalline silica can also be used) at room temperature in unbuffered (pH<7) NaCl solutions of high ionic strength (Fig. 1). Silica can be used as a simple additive, sedimented in a microcentrifuge tube to separate it from the unbound species left in solution, or it can be used in a mini spin column with a porous disk support to facilitate washing and elution steps. The supporting disk pore size should be smaller than the silica particle size (e.g., we have employed a spin column with a 20 μm pore polyethylene filter for 40-63 μm silica particles). At least 98% of HA (9-1700 kDa) applied in 5 M NaCl is bound at loads of 1-10 μg HA per 140 μg of silica in a mini spin column format. HA binding is size-independent in 4-6 M NaCl, whereas in 2.5-3 M NaCl, the low M HA is incompletely bound, and binding for all sizes of HA is weak in ≤2 M NaCl. The preferred pH of the solution is pH 5.5-7.0, to reduce deprotonation of silanol groups and minimize repulsion between silica and polyanionic HA. (In unbuffered aqueous solutions, absorption of CO₂ from the air leads to formation of H₂CO₃, dissociating to H⁺ and HCO₃⁻, such that the solution pH tends toward a pH of ca. 5.5-6.0 over time.)

Figure 1. HA binds silica reversibly in concentrated NaCl solutions.

Six identical HA samples consisting of a mixture of four HA standards (1700 kDa, 150 kDa, 50 kDa, 9 kDa, 0.5 μg each except 1 μg for 9 kDa HA) were loaded on six silica columns (140 μg silica each) at NaCl concentrations of 0-4 M. After binding HA, each silica column was washed in 5 mM Tris - 0.5 mM EDTA (pH 7.1) containing 50% trifluoroethanol, then HA was eluted in a three step procedure using 1) 10 mM Tris - 1 mM EDTA (pH 7.5), 2) PBS (pH 7.4), and 3) 0.4 M KOH. The six paired unbound and bound+eluted fractions were desalted, dried and dissolved in water. A) Polyacrylamide gel electrophoresis on 4-20% polyacrylamide with Stains-All detection. BPB = bromophenol blue tracking dye. B) HA concentration assayed by AlphaScreen.

HA also binds to powdered silica in 3-5 M NaI solutions with similar M dependence. HA of all sizes binds weakly to silica in 5 M ammonium acetate. No HA binding to silica was observed in 6 M GuHCl (pH 3-5.5). We did not observe binding of HA to the powdered silica or silica-based membranes or fibers in any tested commercial kit for DNA or RNA isolation using chaotropic GuSCN or GuHCl solutions for the binding step, in the presence or absence of added ethanol.

Other non-silica media were tested for HA binding and release. HA does not bind to endotoxin removal columns containing ε-poly-L-lysine-modified cellulose beads, in water or in phosphate buffered saline (PBS) at 25-100% strength. HA will bind poly-L-lysine-modified agarose in 10-50% strength PBS, but the binding is strongly M dependent (preferentially binding high M HA) and difficult to release.

Washing silica-bound HA to remove salts and contaminants can be accomplished at room temperature with a low ionic strength solution such as 5 mM Tris- 0.5 mM EDTA at pH 7.1, containing 50% trifluoroethanol. Lower trifluoroethanol concentration results in release of HA during the washing step. Replacing 50% trifluoroethanol with 75% ethanol in Tris-EDTA maintains similar overall retention of HA on silica, but with some loss of the smallest HA sizes tested (9-50 kDa).

Elution of HA of low to moderate M (up to ca. 150 kDa) from silica can be performed at room temperature using low ionic strength 10 mM Tris-1 mM EDTA buffer at a slightly higher pH of 7.5, followed by a second elution step using phosphate-buffered saline (PBS) at pH 7.4. Elution of high M HA may be incomplete after the first two steps, but can be achieved by a third overnight step using a basic solution such as 0.1-0.4 M NaOH or KOH, preferably at 4°C, followed by neutralization. No degradation of HA during binding or elution at room temperature or at 4°C was observed, and the final yield of HA eluted after binding in 2.5-4 M NaCl was over 70% (Fig. 1).

The molecular weight distribution of an HA sample mixture loaded on silica in 5 M NaCl and then eluted by the three-step procedure was quantitatively compared with that of the untreated control HA mixture. Densitometric analysis of the stained PAGE gel showed that the eluted HA profile closely resembled that of the original sample (Fig. 2)

Figure 2.

HA size distribution is maintained following adsorption to silica in 5 M NaCl and elution. A) 4-20% polyacrylamide gel image with Stains-All detection: lane 1, HA eluted after adsorption to silica in 5 M NaCl; lane 2: untreated control HA mixture. The HA mixtures contained 0.5 μg each of 1700, 150, and 50 kDa HA, and 1 μg of 9 kDa HA. B) densitometric analysis of lanes 1 and 2, with relative mobility of 1.0 corresponding to the marker dye bromophenol blue. The profiles are offset vertically to facilitate comparison.

3.2. HA can be separated from sulfated GAGs by selective adsorption to silica.

A mixture of pure HA and sulfated GAGs can be fractionated by adsorption to silica. Binding of chondroitin sulfate (CS), heparan sulfate (HS), and heparin (HP) to silica has not been detected at any NaCl concentration from 2-5 M. Fig. 3 shows that HA binding to silica in 5 M NaCl allows it to be separated from sulfated GAGs. Additional supporting data provided as Supplementary Figures S2 and S3 show that densitometric analysis can be used to establish the lack of significant sulfated GAG contamination.

Figure 3. HA can be separated from sulfated GAGs by selective adsorption to silica.

Electrophoretic analysis on 4-20% polyacrylamide of pure GAG standards (Lanes 1-4), a mixture of the GAGs (Lane 5) and HA purified from the mixture by adsorption to silica in 5 M NaCl and elution using 10 mM Tris - 1 mM EDTA (pH 7.5), PBS (pH 7.4), and 0.1 M NaOH (Lane 6). Stains-All detection. Lane 1: HA (top to bottom): 1700 kDa, 150 kDa, 50 kDa, 31 kDa, 1 μg each. Lane 2: Chondroitin sulfate, ca. 40 kDa, 2 μg. Lane 3: Heparan sulfate, ca. 40 kDa 2 μg. Lane 4: Heparin, ca. 16 kDa 1 μg. BPB = bromophenol blue tracking dye.

Use color for online version; gray for print

Because RNA binding to silica can be enhanced by addition of ethanol, we tested GAG binding to silica in 2.5 M NaCl containing 10-30% ethanol, but found no HA or sulfated GAGs bound to silica in those conditions. Other solvents were tested to seek conditions under which sulfated GAGs could bind silica, but none have been identified.

3.3. HA isolation from conditioned medium of cultured cells.

HA spiked (with sulfated GAGs) into conditioned medium obtained from HEK293 cells can be isolated by proteolytic digestion, which leaves the GAGs intact, followed by adsorption to silica in 5 M NaCl to obtain pure HA (Fig. 4). (HA can also be purified from conditioned medium without protease treatment, but the yield is reduced.) Note that the HA bands retain their nearly monodisperse sizes and are not degraded by the isolation procedure. There is a low degree of contamination of the HA by an unidentified fast-migrating blue-staining species. We tentatively ascribe that species to a protein that is incompletely degraded by proteinase K and that binds silica.

Figure 4. Polyacrylamide gel electrophoresis demonstrating HA purification from HEK293 cell conditioned medium spiked with standard HA samples and sulfated GAGs.

All samples in lanes 7-10 were desalted on a ZebaSpin column, dried by centrifugal evaporation, and re-dissolved in water before analysis, with 80% of each sample loaded on the gel. Lane 1: Mixture of pure HA standards of differing sizes. From top to bottom: 1700 kDa, 150 kDa, 50 kDa, 31 kDa, 1 μg each. Lane 2: Chondroitin sulfate standard, 2 μg. Lane 3: Heparan sulfate standard, 2 μg. Lane 4: Heparin standard, 1 μg. Lane 5: Mixture of HA and sulfated GAG standards. Lane 6: blank. Lane 7: Conditioned medium, untreated, 250 μl original volume. Lane 8: Conditioned medium, 250 μl original volume, spiked with HA + sulfated GAG standards. Lane 9: Conditioned medium, 250 μl original volume, spiked with HA + sGAGs, then subjected to proteinase K digestion. Lane 10: Conditioned medium, 250 μl original volume, spiked with HA + sGAGs, then proteinase K digestion and purification of HA using adsorption to silica in 5 M NaCl and elution using 10 mM Tris - 1 mM EDTA (pH 7.5), PBS (pH 7.4), and 0.4 M KOH.

Use color for online version; gray for print

3.4. HA isolation from human serum.

HA spiked into human serum can be isolated using silica adsorption, after removal of most protein using an albumin affinity column, with or without subsequent proteolytic digestion (Fig. 5). Note that the HA bands retain their nearly monodisperse sizes and are not degraded by the isolation procedures. Because the high albumin concentration in serum can reduce HA binding to silica, an affinity column containing the dye Cibacron Blue was employed to reduce the protein content (compare lanes 2 and 3 in Fig. 5). Following serum albumin removal, HA can be isolated directly using silica adsorption (lane 6), but with reduced yield, as a result of remaining protein interfering with HA adsorption. Higher yield of HA is obtained by using proteinase K treatment after the albumin removal (lane 4), then followed by silica adsorption and elution (lane 5). Comparing lanes 5 and 6 (with or without proteinase K treatment before silica), the yield of HA is greater after proteinase K (lane 5), but a fast-moving blue band is again observed as a contaminant in proteinase K-treated samples. Without proteinase K (lane 6), there is a contaminant band moving approximately halfway down the gel, as observed for the conditioned medium samples.

Figure 5. Polyacrylamide gel electrophoresis demonstrating HA purification from 100 μL human serum samples spiked with HA standards ranging in size from 9-1700 kDa.

All samples in lanes 2-6 contained 100 μl original serum volume and spiked HA standards. For lanes 5 and 6, for the final silica step samples were brought to 0.5 M NaCl, then adsorbed to silica and eluted following the standard procedure. All samples in lanes 2-6 were desalted on a ZebaSpin column, dried by centrifugal evaporation, and re-dissolved in 10 μl water before electrophoretic analysis. Lane 1: Mixture of pure HA standards of differing sizes. From top to bottom: 1700 kDa, 150 kDa, 50 kDa, (0.5 μg each), and 9 kDa (1 μg). Lane 2: Untreated serum. Lane 3: Serum treated with Affi-Gel albumin affinity column. Lane 4: Serum treated with Affi-Gel and proteinase K. Lane 5: Serum treated with Affi-Gel, proteinase K, and silica. Lane 6: Serum treated with Affi-Gel and silica.

Use color for online version; gray for print

3.5. HA can be separated from other unsulfated GAGs by choice of salt concentration for adsorption to silica.

Unsulfated GAGs chondroitin and heparosan were found to bind silica in 1.7-4 M NaCl. Because HA binds silica less well at low salt concentrations, chondroitin and heparosan can be separated from HA by selectively adsorbing those GAGs to silica in 2 M NaCl (Fig. 6). For routine HA isolation from biological samples with low chondroitin or heparosan content, a pretreatment or cleanup step using 2 M NaCl binding may be unnecessary.

Figure 6. HA can be separated from chondroitin (CH) and heparosan (HN) by salt concentration-specific adsorption to silica.

A) 4-20% polyacrylamide gel electrophoretic characterization of polydisperse CH and quasi-monodisperse HN samples, 1 μg each. B) Analysis of the unbound fraction of a GAG mixture of HA (5 μg) + HN (1 μg) + CH (1 μg) after loading on silica at the indicated NaCl concentrations shows that HN and CH are removed at lower NaCl concentration than HA. Six different silica columns were loaded at the indicated salt concentrations. C) Analysis of GAGs (mixed HA + HN + CH) retrieved after binding at the indicated NaCl concentrations shows that 2 M NaCl can be used to obtain CH and HN, but binding in 4 M NaCl allows all three types of unsulfated GAGs to be obtained.

3.6. HA can be separated from DNA and RNA by choice of salt concentration for adsorption to silica.

DNA and RNA are well known to bind silica in solutions of chaotropic salts like GuHCl and GuSCN. The adsorption of DNA to silica in approximately 8 M NaI solutions[63] and in NaCl solutions above about 2 M at pH 5.2[65] has also been documented. We confirmed the NaCl concentration dependence of silica adsorption for DNA and RNA ladder samples. DNA restriction fragments (100 bp ladder sample containing DNA of 100-1517 bp) and RNA fragments (ssRNA ladder containing RNA of 500-9000 bases; not shown) were found to bind to silica in unbuffered (pH ca. 5.5-7.0) 2 M NaCl (Fig. 7). Thus DNA and RNA, as with unsulfated GAGs chondroitin and heparosan, can be removed from isolated HA by adsorption to silica in 2 M NaCl.

Figure 7. DNA can be removed from HA by adsorption to silica in 2 M NaCl.

Electrophoretic analysis on 4-20% polyacrylamide of HA and DNA 100 bp ladder. Lane 1: HA (top to bottom: 150 kDa, 50 kDa) 1 μg each. Lane 2: DNA 100 bp ladder standard, 100-1517 bp, 3 μg. Lane 3: Mixture of HA + DNA standards before silica. Lane 4: Unbound sample from HA + DNA exposed to silica in 2 M NaCl, showing that most DNA was removed by adsorption to silica in 2 M NaCl, but HA was not removed.

4. Discussion

The method for isolation of HA by selective control of adsorption to silica and subsequent elution is based on concepts used in isolation of DNA and RNA, but with several surprising differences.

The surface of amorphous silica has siloxane bridges (Si-O-Si) and silanol (Si-OH and Si-O⁻) groups. The silanols can be isolated, geminal (two OH bonded to the same Si), or vicinal [73, 74]. Siloxane bridges provide a more hydrophobic surface aspect, whereas Si-OH act as hydrogen bond donors for water and biopolymers. The pKa for the silanol groups ranges from about 4.5-8.5, with an apparent pKa of about 7.5 in the presence of bound polyanions[64]. The net charge remains negative under the conditions used in binding DNA or HA.

Binding of anionic biopolymers to silica requires minimization of electrostatic repulsion, and is therefore pH and ionic strength dependent. Optimum binding is at pH less than 7. At pH greater than 7, the net repulsion disfavors binding polyanions. Salt concentration can increase binding at a given pH by reducing repulsion.[64]

For DNA, the driving force for silica binding is significantly entropic, and has been described as a type of hydrophobic interaction between DNA bases and regions of the silica surface poor in silanols[75]. In physiological aqueous solutions, the surface of silica is highly hydrated by water molecules, up to several monolayers thick[74]. Chaotropic salts are used to enhance DNA binding by disrupting ordered water at the polyanion and silica interface[64].

In close proximity to the silica surface, DNA becomes hydrogen bonded to silica. Binding of DNA to silica is slightly exothermic, reflecting formation of H bonds[64, 65], and thus binding is favored at lower temperatures. Once bound, DNA can be difficult to elute[67]. Elevated temperature can increase recovery. NaOH has also been used to help elute DNA[66], and has been proposed to cause silica dissolution sufficient to release DNA.

HA adsorption to silica has some similarities to that of DNA but also significant differences. Similarities include increased binding at pH less than 7, and at high ionic strength. Significant differences include: 1) HA does not bind silica in chaotropic salt solutions like GuHCl or GuSCN. 2) The salt dependence of polyanion binding is not correlated with linear charge density and the electrostatic repulsion from silica. DNA has a linear charge density of 6 negative charges per nm, while sulfated GAGs have 2-4 negative charges per nm, and HA, CH, and HN all have only 1 negative charge per nm but binding affinity in salt solutions is CH, HN, DNA > HA ≫sGAGs (i.e. HA needs a higher NaCl concentration to bind silica, despite the much lower linear charge density than DNA and thus weaker repulsion.) 3) Basic pH and elevated temperatures aid elution, but the mechanism is not dissolution of silica. For HA, NaOH increases chain flexibility by ionizing hydroxyl groups and reducing intrachain H bonding[76].

These differences suggest a fundamental difference in binding mechanism for HA or other unsulfated GAGs versus DNA, possibly mediated by ordered water layers between silica and HA, CH, or HN. In addition, fit of the GAG structure to the ordered water layer on silica may be important. On a rough surface of amorphous silica, biomolecules can differ in fit and surface recognition[74]. One hypothesis is that CH and HN have slightly better fit than HA, and therefore require less NaCl to minimize repulsion. In contrast, the failure of sulfated GAGs to bind under any conditions tested suggests a poor fit to the silica surface due to the bulky sulfate groups.

The selective and reversible adsorption of HA of all sizes from dilute solutions to amorphous silica as described here differs from the previously reported deposition of HA onto silica disks using spin coating and drying of concentrated solutions of high M HA in distilled water, forming a thick layer of entangled polymer chains[77, 78].

5. Conclusions

We have developed a simple and rapid procedure for selective isolation of HA from biological samples, while maintaining its native size distribution, ranging at least from approximately 9 kDa to several MDa, which is important to its biological function. The method avoids the use of highly toxic organic materials like phenol or chloroform. Closely related molecules like sulfated glycosaminoglycans and the unsulfated glycosaminoglycan chondroitin are separated from HA by proper choice of solvent ionic strength and composition. The isolated HA so obtained is suitable for analysis of quantity and molecular mass distribution, and may be further fractionated by size using multiple different methods, enabling its use as a biomarker in medical diagnostics. While our current method apparently produces a high quality HA, it is recommended that users test their own preparations (especially vertebrate tissue extracts) with FACE gels or HPLC disaccharide analyses with the appropriate GAG lyases. Towards this approach, it should be noted that the chondroitin ABC lyases can also cleave HA, thus good resolution between the HA and the CS fragments is required especially when trace contaminants are predicted.

Highlights.

• Hyaluronan can be isolated by adsorption to silica at high salt concentration

• Hyaluronan can be separated from sulfated glycosaminoglycans

• Hyaluronan can be separated from unsulfated chondroitin and heparosan

• Hyaluronan can be separated from nucleic acids

• Hyaluronan was isolated from human serum and cultured cell conditioned medium