Planning, evaluating and vetting receptor signaling studies to assess hyaluronan size-dependence and specificity

Abstract

Exciting discoveries in many diverse fields of hyaluronan (HA) biology over the last 40 years have centered around the ability of HA to bind cell surface HA receptors (e.g., CD44, Layilin, LYVE-1, HARE/Stab2 and RHAMM) and sometimes also to activate intracellular signal transduction pathways, frequently involving ERK1/2. Although perplexing, a major characteristic of HA-mediated signal pathway activation for some receptors has been a dependence on the size of the bound HA. Receptors that directly interact with HA, which may not include TLR2/4, bind very well to any HA molecule >8–20 sugars, depending on the receptor. Despite their ability to bind virtually any size HA, only HA chains of a particular mass range can activate receptor-mediated cell signaling. Many studies have demonstrated parts of this emerging story by utilizing different: HA receptors, cell types, animal models, HA sources, HA sizes, assays to assess HA mass and varying controls to verify HA specificity or HA size-dependence. Recent reports have highlighted issues with potential endotoxin contamination of HA fragments, especially those generated by hyaluronidase digestion. Also, researchers unfamiliar with HA polydispersity must adjust to working with, and interpreting data for, preparations without a unique molecular mass (molecular weight). The confusion, uncertainty and skepticism generated by these and other factors has hindered the development of a general consensus about HA-specific and HA-size dependent receptor activation. An overview of issues, suggested strategies and validating controls is presented to aid those planning an HA-mediated receptor signaling study or those trying to evaluate the literature.

Since the discovery by West et al. (1985) that hyaluronan (HA) oligosaccharides can stimulate angiogenesis, many groups have reported the ability of cell surface HA receptors to mediate the activation of intracellular signal transduction pathways (Fitzgerald et al. 2000; Lesley et al. 2000; Bono et al. 2001; Gao et al. 2008; Jiang et al. 2011; Turley et al. 2002; Stern et al. 2006; Pandey et al. 2013). The HA receptors CD44, HARE/Stab2, Layilin, LYVE-1 and RHAMM all bind HA specifically and can initiate signal transduction depending on the circumstances (e.g., cell type or HA size). TLR2 and TLR4 are clearly involved in HA-mediated signal transduction but they may not actually be HA receptors, since their ability to bind HA directly has been assumed, but not demonstrated, and does not occur in at least one binding assay (noted in Weigel and Baggenstoss 2017). In addition to possible inherent differences in signaling characteristics among the six major HA receptors, investigations of the same receptor can yield diverse results in different cell types or animal models. The possibility that signaling by hyaluronidase-generated HA fragments can be due to endotoxin contamination has also raised questions about studies with HA fragments (Huang et al. 2014; Dong et al. 2016). Two less appreciated issues in assessing HA signaling studies, especially for investigators with little experience handling HA, relate to the difficulty of handling large (>MDa) HA without breaking it and to the polydispersity of HA. An HA preparation has no true molecular weight, but rather it has a broad range of sizes (up to 50-fold between the smallest and largest species present) and its size is characterized as either a number-average or weight-average molar mass (Weigel and Baggenstoss 2017). Thus, a major area of confusion in trying to compare signaling studies has centered on the HA itself: the sources, sizes, assays to assess HA mass and controls to verify HA specificity and size-dependence. This article presents some well-defined variables under investigator control and specific control experiments that should be considered when either planning, performing or reviewing any study of HA-mediated receptor signaling.

Investigators have five important parameters under their control for demonstrating the specificity and size-dependence of HA signaling: (1) HA source; (2) HA size; (3) HA purity after in-lab treatments; (4) verification that reagents added to cells are free of, or very low in, lipopolysaccharide (LPS) and (5) controls to verify that any response is truly HA-specific. Mammalian cells are hyper-sensitive to endotoxins, which are LPS fragments derived from the outer wall of Gram^- bacteria, including Escherichia coli. For example, TLR4 is activated at pM levels of LPS (Wakelin et al. 2006), which is found everywhere in our environment including drinking water and even deionized water (Di Luzio and Friedmann 1973). Therefore, it is a daunting challenge for any company or lab to prepare LPS-free reagents and then to keep these special reagents free from LPS contamination during their use and storage.

HA source

Animal-derived HA should not be used. For two important reasons, the gold standard for signaling studies is medical grade HA made by bacteria.

1. HA for human use is no longer made from animal sources, although Healon (rooster comb HA) was developed for eye surgery in 1980 by Pharmacia as the first medical grade HA product. Currently world-wide, large scale commercial HA production for cosmetic or clinical use (as medical devices) is by fermentation of streptococcal or other bacteria expressing native or recombinant HA synthase. Medical grade products are the highest purity and essentially LPS-free in order to meet strict regulatory requirements for their production and clinical use. In contrast, nonmedical-grade HA from any biochemical sales vendor, even with evidence of purity, should be assumed to be contaminated by LPS (Di Luzio and Friedmann 1973; Wakelin et al. 2006; Dong et al. 2016) and should be tested in-house. Highly purified HA can still contain LPS, which is potent at pM concentrations.

2. Animal-derived HA is a poor choice for signaling studies, since it can be contaminated by hundreds of small potent signaling molecules present in animal tissues (e.g., cytokines, hormones, growth factors). For example, partially purified umbilical cord HA, which has been used in many signaling studies, contains up to 2% protein by weight. This high level of contamination in a 1 mg/mL HA solution means that 2 μM protein (average 10 kDa MW) is present and the potential for non-HA-mediated signaling is high. These mammalian contaminants are not present in HA made by fermentation processes.

HA size

Almost all of the molecules used by biomedical researchers (e.g., proteins and DNA genes) are characterized by exact and known formula weights. Thus, it may be difficult to appreciate the importance of understanding the concepts of size distributions, polydispersity and the fact that a product label of “X” kDa HA does not indicate a molecular weight of “X.” It is important that users determine the range of sizes in any HA preparations they use.

1. Methods to determine HA size ranges. Assessing the range of sizes present in a sample is usually more meaningful than estimating an average HA size. Among many possible methods, size-exclusion chromatography coupled to multiangle light scattering (SEC-MALS) has an advantage in that standards are not needed. Many groups do not have access to an SEC-MALS instrument, although these are often associated with crystallography groups. Agarose gel electrophoresis is an excellent way for most labs to assess the average size and the size distribution range for an HA sample (Cowman et al. 2011; Weigel et al. 2013), if narrow-size range standards are available. Most biomedical scientists, with limited access to more technical or expensive biophysical instrumentation, can perform electrophoretic analyses. To help these investigators, the Cowman group developed reproducible protocols for the size analysis of large MDa HA by agarose gel electrophoresis (Cowman et al. 2011) and of 5–500 kDa HA by combined agarose and polyacrylamide electrophoresis (Bhilocha et al. 2011). Narrow-size range HA preparations made in-house (e.g., Fig. 1 in Weigel and Baggenstoss 2017) can be used as standards once their masses have been determined relative to very-narrow-size range standards.

2. Variability. Different techniques often yield different size values for the same HA sample and the method by which the HA mass of a product was determined, if done at all, may not be noted. For a researcher using HA from various sources, therefore, it could be very important to determine the average size and the size distribution of the HA preparations to be used by performing mass analyses of all HAs in their lab at the same time. It is especially important in interpreting results to know the degree of size overlap among different HA preparations.

HA purity

1. If nonmedical-grade HA is used it will most likely not be pure. In-lab manipulations of medical grade HA and production of HA fragments using hyaluronidase will necessarily introduce contaminants. Streptomyces lyase and testicular hyaluronidase, most commonly used by researchers to prepare smaller HA fragments, are crude preparations contaminated by hundreds of host cell components and, more significantly for signaling studies, they contain LPS (Dong et al. 2016). No recombinant or affinity-purified products are available for these two HA degrading enzymes, although many recombinant proteins made in bacteria still contain LPS (Wakelin et al. 2006). Until LPS-free testicular hyaluronidase or Streptomyces lyase is available, a better choice for maintaining low LPS during production of small HA oligosaccharide fragments (e.g., for use in HA-size-specific controls) is to use recombinant PH20 hyaluronidase made for medical use and free of LPS (Huang et al. 2014). Also known as SPAM1, it is currently available from MyBioSource and R&D Systems; always verify that the product ordered has low or no LPS.

2. If larger HA fragments must be made: most importantly avoid using testicular hyaluronidase or Streptomyces lyase and use vetted protocols that monitor and minimize LPS [Dr. Edward N. Harris (Department of Biochemistry, University of Nebraska at Lincoln) can be contacted for information regarding protocols for the production of low-endotoxin HA fragments as described in Pandey et al. (2013). Email: eharris5@unl.edu]. Instead, either: (a) purify the desired sizes of HA from a Lifecore HA lot already near the target mass or (b) treat medical grade MDa HA with mild acid and purify the HA of the desired mass range.

   1. Lifecore Biomedical produces medical grade HA in a wide range of sizes useful to the HA research community. Their processes for generating smaller HA size ranges do not involve chemical or enzymatic treatment. In our experience, HA near the signaling size desired usually has no or low bioactivity, due to inhibition by the nonsignaling HA sizes present (Pandey et al. 2013). Significantly, SEC fractionation, pooling and processing of sizes that are bioactive as well as inactive sizes yields both the desired active-size HA as well as HA sizes that do not signal; importantly, these latter identically processed inactive HA preparation are excellent negative controls.
   2. If the starting material is medical grade MDa, then one can also generate fragments by treating HA with mild HCl under conditions that cleave the GlcNAcβ(1,4)GlcUA glycoside bonds but not N-acetyl bonds; 50–100 mM HCl, 52–55°C (McGary et al. 2003; Pandey et al. 2013). Heat treat for 1–4 h until the average size is in the desired target range and, after neutralization, fractionate the hydrolyzed HA by SEC to prepare narrow-size range fractions; use protocols that minimize or prevent LPS contamination [Dr. Edward N. Harris (Department of Biochemistry, University of Nebraska at Lincoln) can be contacted for information regarding protocols for the production of low-endotoxin HA fragments as described in Pandey et al. (2013). Email: eharris5@unl.edu].

3. If lyase or testicular hyaluronidase treatment is used, then HA fragments should be purified by size exclusion or ion exchange chromatography and LPS should be removed from the starting or purified material, which may not be possible, by use of a commercial resin (e.g., polymyxin B) or preferably Triton-X114 treatment (Liu et al. 1997).

4. Assess purity. A simple crude assessment of purity is to measure the A₂₁₀, A₂₆₀ and A₂₈₀ values of the HA product or HA fragments, dissolved directly in water or NaCl (no acetate, carbonate or phosphate buffers). N-Acetyl groups [–NH(C=O)–CH₃) of GlcNAc absorb at 210 nm, so HA has a high extinction coefficient. Since HA lacks aromatic groups, it shows no absorbance at 260 nm or 280 nm. If samples contain nucleic acid or protein contaminants, however, the A₂₆₀ and A₂₈₀ values, respectively, will reflect their presence and relative amounts (e.g., Fig. 1A in Weigel et al. 2017). The absence of A₂₆₀ and A₂₈₀ signals does not necessarily mean the sample is pure, since contaminants could be present at levels too low for detection (note that HA samples may be LPS-contaminated after exposure to multiple surfaces for this test). A more sensitive method for assessing protein contaminants is amino acid analysis. Another simple method to assess HA purity is to use Stains-All when developing agarose gels because HA, proteins, and nucleic acids have different colors when stained (Green 1975); HA and DNA are blue, protein is red and RNA is bluish-purple.

Verify that anything added to cells has very low or no LPS

Perform LPS testing routinely on all in-lab solutions added to the HA used or to cells. In addition to being present in almost all nonmedical-grade products, LPS is very difficult to remove and even then, if that is successful, it is often accompanied by substantial loss of the desired component. Therefore, preventing LPS introduction into lab reagents is critical. An example is the deionized distilled water every lab uses daily; it may immediately come out LPS-free but how often is the storage container or tubing cleaned or is the tubing end (an air–water interface) checked for slime (microbes)? The widely used standard Limulus Amebocyte Lysate assay is sensitive, easy to use and is available in a range of formats from companies such as Lonza, GenScript and Charles River. Investigators should verify and note in publications the low LPS levels in the various solutions added to cells. However, even with extensive testing for LPS and utilizing only low LPS reagents, one cannot conclude that the observed signaling with a low LPS HA preparation is actually due to the HA.

Controls should verify that a signaling response is specific for HA

The central concern is an artifact; that an HA sample may activate signaling due to non-HA contaminants such as LPS, hormones, cytokines or other bioactive peptides or small molecules. Therefore, negative controls are critical and one cannot perform too many controls to verify that the observed response is due to HA and not to contaminants. Reviewers should only accept studies that convincingly show that the observed HA signaling is due to HA. As many of the following controls as possible should be included in any signaling study.

1. HA samples without signaling ability strongly support conclusions for both HA-specific and size-specific signaling, but only if all samples are prepared in exactly the same manner with the same components present as in the active HA sample. LPS testing of an HA sample alone is not enough if it is diluted into a buffer or anything is added to it before addition to cells or if it was treated with hyaluronidase. A confounding issue with LPS is that its presence at the low levels needed for activating cells is independent of either product purity or sterility; autoclaved items can still contain active LPS.

2. Cells identical to or derived from the cells being tested, which do not express the target receptor or that express a receptor mutant impaired in HA binding or signaling, provide a very good negative control to assess signaling mediated by a non-HA-dependent process (or the possibility that a second HA binding protein is involved in the signaling). If receptor-negative cells are not available, then a receptor knock-down experiment (e.g., using siRNA) should be shown to demonstrate impaired or eliminated signaling in response to the HA used. A highly specific and vetted monoclonal antibody that blocks HA binding to the receptor could also provide a validating control.

3. Receptor signaling in response to increasing HA concentration should be assessed to determine if the dose response curve is biphasic or hyperbolic. HA-mediated signaling by CD44, HARE/Stab2 and LYVE-1 appears to be associated with receptor clustering or oligomerization, indicating that receptor activation is via a physical-spatial mechanism rather than a direct transmembrane mediated mechanism for receptor cytoplasmic domain activation (Weigel and Baggenstoss 2017). Physical-spatial mechanisms create receptor•HA•[receptor]_n complexes and the activation dose response curve will necessarily be biphasic. Higher concentrations of activating HA will inhibit activation because all receptors are then bound to different HA chains; they can no longer be clustered and brought into close proximity since they are not bound to the same HA molecule. For example, HARE-mediated ERK1/2 activation occurs in a biphasic way with a maximal response at about 10 nM HA (Kyosseva et al. 2008). Finding a hyperbolic response, in which no inhibition occurs at higher HA concentration does not mean that signaling is not mediated by the HA, but investigators and reviewers should be suspicious and require additional supporting controls for HA specificity.

4. A mixing-blocking control with nonsignaling HA could reveal non-HA-dependent signal activation. HA fragments of increasing size bind increasingly well and HA fragments >20 sugars bind strongly to any HA receptor. Therefore, excess nonsignaling HA will naturally displace signaling HA and inhibit signaling (Pandey et al. 2013). If a mixture of the active signaling HA and excess inactive HA still activates signaling, then the signaling of the “active HA” cannot be due to HA. If this conclusion is not accepted by the investigators, then they would need to demonstrate that the nonsignaling HA, in the mixture of HA preparations, is not bound by the receptor, which would be a very remarkable discovery.

Summary

Although the above discussion of issues and suggestions is not exhaustive, surprisingly few published studies actually address or satisfy most of the criteria presented. This does not mean the authors did not consider or perform some of the analyses or controls noted, but they may not have been reported. Typically, the weakest area of many studies is a lack of sufficient controls to validate that the observed responses are specific to the receptor and to the HA and not due to contaminating LPS, hormones, cytokines or other bioactive molecules.

Abbreviations

HA, hyaluronic acid or hyaluronate or hyaluronan; LPS, lipopolysaccharide (endotoxin).

Funding

The supporting research for this article was funded in part by National Institutes of Health grants GM035978 and GM069961 and by the Ed Miller Chair in Molecular Biology.

Conflict of interest statement

None declared.