A single molecule assay to probe monovalent and multivalent bonds between hyaluronan and its key leukocyte receptor CD44 under force

Glycosaminoglycans (GAGs), a category of linear, anionic polysaccharides, are ubiquitous in the extracellular space, and important extrinsic regulators of cell function. Despite the recognized significance of mechanical stimuli in cellular communication, however, only few single molecule methods are currently available to study how monovalent and multivalent GAG·protein bonds respond to directed mechanical forces. Here, we have devised such a method, by combining purpose-designed surfaces that afford immobilization of GAGs and receptors at controlled nanoscale organizations with single molecule force spectroscopy (SMFS). We apply the method to study the interaction of the GAG polymer hyaluronan (HA) with CD44, its receptor in vascular endothelium. Individual bonds between HA and CD44 are remarkably resistant to rupture under force in comparison to their low binding affinity. Multiple bonds along a single HA chain rupture sequentially and independently under load. We also demonstrate how strong non-covalent bonds, which are versatile for controlled protein and GAG immobilization, can be effectively used as molecular anchors in SMFS. We thus establish a versatile method for analyzing the nanomechanics of GAG·protein interactions at the level of single GAG chains, which provides new molecular-level insight into the role of mechanical forces in the assembly and function of GAG-rich extracellular matrices.

in working buffer. The complex contains multiple biotin moieties at random (uncontrolled) positions.

Quantification of CD44 surface density by spectroscopic ellipsometry (SE):
To estimate the surface density of CD44, the formation of CD44 monolayers was followed by SE. SE measurements were carried out at room temperature with a M2000V system (J. A. Woollam, Lincoln, NE, USA) and the data were analyzed using CompleteEASE software (J. A. Woollam) following established procedures. 3,4 Briefly, gold-coated silicon wafers functionalized with biotinylated OEG monolayers were installed in a custom-built open cuvette (~180 μl volume, passivated with 10 mg/ml bovine serum albumin for 20 min prior to use). The cuvette featured a magnetic stirrer which was used to homogenize the cuvette content for 20 s after injecting the sample into the cuvette and during rinses with working buffer; sample binding was followed in still solution. In the optical model to fit the data, the opaque gold film and the OEG monolayer were treated as a single isotropic layer and fitted as a B-Spline substrate, and the protein film made of SAv and CD44 was treated as an isotropic and transparent Cauchy layer. Areal protein mass densities were determined through de Feijter's equation, 5 using a refractive index increment of 0.18 cm 3 /g. 6,7

AFM SMFS of SAv•biotin and TAv•biotin interactions:
Gold-coated AFM cantilevers were conditioned by exposure to UV/ozone for 30 min, and then immersed overnight at room temperature in a solution of linear OEG (7 units) with a hydroxyl and a thiol group at the ends (OEG thiol; Polypure) mixed with linear PEG (10 kDa) with a biotin and a thiol group at the ends (b-PEG thiol; IRIS biotech GmbH, Germany) in ultrapure water (total concentration 1 mM, molar ratio 4⨯10 4 OEG per 1 b-PEG). The substrates were then rinsed in ultrapure water, and immersed in working buffer for their final use.
The mixed OEG/b-PEG monolayer thus formed displayed biotin at a dilution that showed a satisfactory binding frequency in AFM force spectroscopy on planar substrates displaying monolayers of SAv or TAv. Specifically, 65% of all force curves showed no specific unbinding event and 35% showed one event; two or more events were not observed. Monolayers of SAv and TAv were prepared and force data at various retract velocities acquired as described in the Methods section in the main text. Force curves were fit (JPK Data Processing Software) with a freely jointed chain (FJC) model, with the Kuhn segment length fixed to 0.7 nm (which describes the stretching of PEG chains in aqueous solution well over the force range relevant to our assays 8 ) and the contour length as the only adjustable parameter. The effective spring constant eff , corresponding to the slope of the best-fit FJC model curve close to the rupture point, was used to compute the instantaneous loading rate = eff .

SUPPLEMENTARY FIGURES
Supplementary Figure S1. QCM-D immobilization assays for end-biotinylated HA (840 kDa, 2.1 μm contour length) on biotinylated OEG monolayers, presented analogous to Fig. 1. The data in (a) confirm formation of a stable and dense SAv monolayer of about 4 nm thickness, 9,10 and subsequent grafting of HA through biotin at the reducing end. The frequency decrease upon HA incubation was small (∆f = -1.3 ± 0.5 Hz) and the concomitant dissipation shift large (∆D = 3.1 ± 0.2 × 10 -6 ), consistent with previously reported data, and reflecting the formation of a very soft and hydrated film. 10,11 HA polymers lacking biotin do not bind to SAv monolayers, as shown previously. 10 Data in (b) demonstrate that the CD44 construct employed in this study does not bind to a SAv monolayer that had previously been saturated with soluble biotin. Exposure to biotin does not induce a measurable QCM-D response due to biotin's small size (cf. Fig. S3b), and this trace is not shown here. AFM probes were functionalized with HA as established through the QCM-D assay, but with a reduced HA incubation time of 6 min. At physiological ionic strength, an unperturbed HA chain of 2.1 μm contour length is expected to form a random coil with a radius of gyration of approximately 75 nm. 12 Considering the geometry of the AFM probe and cantilever, and assuming mass-transport limited binding of HA, we can estimate an upper limit of 24 ng/cm 2 for the HA surface coverage, corresponding to a root-mean-square distance between grafting points of 76 nm. 13 This implies that the HA film is in the so-called 'mushroom regime', i.e. immobilized polymer coils barely interpenetrate and largely retain their random-coil conformation. 14 Because of the sharpness of the AFM probe (the apex radius is typically 30 nm; cf. Fig. 1a) and the grafting density and conformation of HA, we expect only one or at most very few HA molecules to be able to contact the protein-covered surface. Figure S2. QCM-D immobilization assays for CD44 on SLBs, presented analogous to Fig. 1. SLBs were formed by spreading of small unilamellar vesicles (SUVs) composed of a DOPC:(NTA) 3 -SOA mixture (molar ratio 19:1); the two-phase responses in Δf and ΔD, the small final dissipation shift (˂ 0.3 × 10 -6 ) and the final frequency shift (-29 ± 1 Hz) are characteristic for the formation of SLBs of good quality. 4,15 QCM-D responses in (a) indicate the formation of a stable and HA-binding CD44 monolayer, and demonstrate that CD44 is specifically immobilized through its polyhistidine tag to the Ni 2+ -NTA moieties in the SLB (i.e. it can be fully eluted in imidazole). Frequency and dissipation shifts at about 120 min reflect the changes in the viscosity and density of the solution due to the presence of imidazole and are not due to surface effects. Data in (b) demonstrate that HA binds through the authentic HA-binding site on CD44 (largely blocked with anti-CD44 Ab). Figure S3. Quantification of CD44 receptor density. Experimental data are shown in black, and the red curve is a best-fit theoretical curve for mass-transport limited binding. Immobilization of CD44 via its C-terminal biotin tag on a SAv-covered OEG monolayer ( Supplementary Fig. S9a-c) was followed by SE to determine the areal protein mass density Γ (see Supplementary Methods for details). The incubation conditions were identical to those used for the preparation of 'high receptor density' samples for AFM force spectroscopy, i.e. 6.5 µg/ml CD44 in still solution for 30 min (here from 6 to 36 min). At sufficiently low coverage, binding is expected to be mass-transport limited because the biotin•SAv bond forms fast. For mass-transport limited binding from still solution, binding is expected to scale with the square root of incubation time, with Γ = 2 b � ⁄ , where c b and D are the adsorbate's concentration and diffusion constant in the bulk solution, respectively. 16 This is indeed the case for CD44 up to Γ ≈ 80 ng/cm 2 : the red curve corresponds to Γ = Γ 0 + �( − 0 ) with = 2 b � ⁄ = 22.8 ± 0.2 ng/(cm 2 min 1/2 ) and where the small offset Γ 0 = 3.5 ± 0.7 ng/cm 2 at t 0 = 5.4 ± 0.1 min arises from the brief initial stirring required at the start of incubation for solution homogenization (errors represent confidence intervals from curve fitting). The progressive reduction in binding rate compared to the red curve at Γ > 80 ng/cm 2 indicates that surface crowding limits binding at high surface coverage. The surface coverage after 30 min of incubation was 110 ± 5 ng/cm 2 (mean ± variations around the mean from two independent measurements). From SDS PAGE analysis, we estimate the molecular weight of CD44 at 60 kDa, and 'high receptor density' surfaces thus correspond to a surface density of 1.8 pmol/cm 2 and an rms distance of 10 nm. The 'low receptor density' surfaces were obtained at 26-fold lower CD44 concentration (0.25 µg/ml) under otherwise identical incubation conditions. Based on the above equations we predict an areal mass density of 4.2 ng/cm 2 corresponding to a surface density of 0.07 pmol/cm 2 and an rms distance of 50 nm. We expect these estimates to be accurate also for CD44 anchorage via polyhistidine tags because binding rates are very similar (compare Fig. 2a with Supplementary Fig. S9a). Figure S4. Oriented protein immobilization is important for protein functionality. (a-b) QCM-D immobilization assay for a complex of aggrecan G1 domain and cartilage link protein (AG1-LP), presented analogous to Fig. 1. In contrast to CD44 and LYVE-1, AG1-LP was tagged using a procedure that is not site-specific; the tag was biotin and immobilization was performed as for HA (cf. Supplementary Fig. S1). Responses upon incubation of AG1-LP indicate stable and specific (i.e. largely blockable by saturation of SAv with free biotin) immobilization of the protein through biotin in a monolayer of 5 nm thickness. HA polymer bound stably and specifically to AG1-LP. Compared to CD44 (Fig. 2), the shifts in frequency (-1.7 ± 0.9 Hz) and dissipation (0.5 ± 0.2 × 10 -6 ) for HA-binding to AG1-LP were small even though AG1-LP is known to bind HA with high affinity. [17][18][19] The reduced HA binding indicates that only a small fraction of the immobilized AG1-LP is active. Most likely, this is the consequence of tagging AG1-LP at random positions, leading to immobilization with an orientation that renders the HA-binding site of a large fraction of the AG1-LPs inaccessible. (c) Representative force-separation curve (pink -approach, black -retract; approach/retract velocity 1000 nm/s) recorded at maximal AG1-LP surface coverage. The curve shows a single rupture event. The red line is a best-fit WLC model curve with a persistence length of 4.1 nm characteristic for HA (cf. Fig. 3). This behavior was observed in 10% of the force curves; 88% showed no rupture event, and 2% showed two or more distinct rupture events. No rupture events (in n = 200 force curves per condition) were observed when either HA on the probe or AG1-LP on the surface were lacking, confirming that the observed interactions are specific. The frequency of binding events is remarkably small, if one considers that the employed HA polymers are long enough to bind about 100 AG1-LPs simultaneously, and that the polymer coil can readily explore a surface area containing tens of AG1-LPs. It is consistent with the limited activity of immobilized AG1-LP observed by QCM-D, and illustrates the importance of ensuring precise protein orientation. Figure S5. Supplementary force spectroscopy data for HA•CD44 interactions. (a) Set of 5 randomly selected force-separation curves (retract velocity 1000 nm/s; conditions as in Fig. 3) featuring a single specific rupture event at various separation distances. The extension curves overlap when normalized, 20 confirming that interactions between a single HA chain and CD44 are probed. (b) Selected force-separation curves, registered at a retract velocity of 2000 nm/s, for control conditions as indicated. No specific rupture events are observed, and this feature is representative for all force curves acquired for the control conditions (n = 200 each). Figure S6. QCM-D immobilization assays for biotinylated HA on TAv monolayers, presented analogous to Fig. 1. The QCM-D responses after TAv binding, and upon subsequent incubations with biotinylated HA (a) are virtually identical to those observed for SAv (cf. Supplementary Fig. S1a), confirming that the switch from SAv to TAv does not alter the surface density and organization of HA. The lack of binding for biotin-free HA (b) confirms specific immobilization through biotin. TAv was incubated sequentially at two distinct concentrations (first at 1 µg/mL and then at 20 µg/mL) for technical reasons; this does not affect the properties of the final TAv monolayer. Figure S7. Immobilization of HA through TAv instead of SAv does not affect the CD44•HA force spectroscopy data. Force spectroscopy data shown here were obtained with HA immobilized via TAv and high CD44 surface density (rms inter-CD44-distance ~10 nm) on a His-tag-capturing sensor. (a) Representative force-separation curve (pink -approach, greenretract; approach/retract velocity 1000 nm/s). The red lines are best-fit WLC model curves (L p = 4.1 nm fixed). Histograms of effective contour lengths (b) and rupture forces (c), and dynamic force spectra (d) are displayed analogous to Fig. 3c-e. Mean rupture forces and standard deviations are virtually identical to those obtained with HA immobilized via SAv (cf. Fig. 4d and Table 1).  Figure S9. Immobilization of CD44 through the biotin tag instead of the polyhistidine tag (both located at the C-terminus) does not affect the force spectroscopy data. (ab) QCM-D immobilization assays, presented analogous to Fig. 1. QCM-D responses in (a) indicate formation of a stable and HA-binding CD44 monolayer on SAv (see Supplementary Fig.  S1b for a demonstration that binding is specific through biotin). Data in (b) demonstrate that HA binds through the authentic HA-binding site on CD44 (largely blocked with anti-CD44 Ab). (c) Representation of the surface architecture (not to scale) displaying CD44 anchored analogous to HA (cf. Fig. 1 and Supplementary Fig. S1) on a SAv-coated biotinylated OEG monolayer. (d-f) Force spectroscopy data obtained at high CD44 surface density (CD44 was incubated at 6.5 μg/ml for 30 min). (d) A representative force-separation curve (pink -approach, green -retract; retract velocity 2000 nm/s). The red lines are best-fit WLC model curves (L p = 4.1 nm fixed). (e) Rupture force histograms displayed analogous to Fig. 3d. (f) Dynamic force spectra displayed analogous to Fig. 3e. Mean rupture forces and standard deviations are virtually identical to those obtained with CD44 immobilized through the polyhistidine tag (cf. Fig. 4d and Table 1).