Javier
Fernandez-Castanon
*a,
Silvio
Bianchi
b,
Filippo
Saglimbeni
b,
Roberto
Di Leonardo
ab and
Francesco
Sciortino
ac
aDipartimento di Fisica, “Sapienza” Università di Roma, Rome, I-00185, Italy. E-mail: javier.fernandez.castanon@roma1.infn.it
bCNR-NANOTEC, Soft and Living Matter Laboratory, Rome, I-00185, Italy
cCNR-ISC, UOS “Sapienza” Università di Roma, Rome, I-00186, Italy
First published on 19th June 2018
We present systematic characterisation by means of dynamic light scattering and particle tracking techniques of the viscosity and of the linear viscoelastic moduli, G′(ω) and G′′(ω), for two different DNA hydrogels. These thermoreversible systems are composed of tetravalent DNA-made nanostars whose sticky sequence is designed to provide controlled interparticle bonding. While the first system forms a gel on cooling, the second one has been programmed to behave as a re-entrant gel, turning again to a fluid solution at low temperature. The frequency-dependent viscous and storage moduli and the viscosity reveal the different viscoelastic behavior of the two DNA hydrogels. Our results show how little variations in the design of the DNA sequences allow tuning of the mechanical response of these biocompatible all-DNA materials.
Most of the literature on rheological properties of DNA solutions has focused on long single and double stranded DNA, interpreted with models of entangled semiflexible polymer coils.22–24 Experimental characterisation of the mechanical properties of materials created starting from purposefully designed DNA-particles is highly welcome and, only very recently, a microrheology study of three-functional Y-DNA particles has been submitted.25 A novel experimental microfluidic-based active microrheology approach to investigate how designed three-functional DNA-hydrogels evolve from diffusive liquid-like systems to shear-thinning fluids on cooling, passing through a series of intermediate power-law fluid states has recently been published.26 Here, we focus on the rheological properties of two DNA hydrogels composed of tetrafunctional DNA nanostars (NSs in the following). In both cases, the NSs are assembled from four properly designed DNA single-strands of about forty bases. At the end of each of the four arms, a short self-complementary single-strand sequence provides a well characterised binding overhang, commonly indicated as a sticky end. Below the melting temperature of the sticky end sequence, the particles bind in a three-dimensional tetravalent network in which each NS acts as a network site. The first NS system is a well characterised all-DNA sample that progressively gels on cooling.10,27 The second system has been designed to display re-entrant gelation28 (the RG system in the following). As for the NS system, the RG system evolves from a fluid at high temperature to a well developed highly connected network in which all DNA NSs are bonded on cooling. This gel then transforms to a fluid phase on further cooling, a behavior that is achieved due to the additional presence in solution of short DNA sequences acting as competitors. Exploiting competitive interactions, these short DNA sequences, at low temperature, contend with the sticky sequences completely capping the NS overhangs. As a result, the system melts both at high and at low temperature.28,29 The self-assembling pathways for both systems as a function of temperature (T) are depicted in Fig. 1.
We report dynamic light scattering (DLS) and particle tracking microrheology (PTM) experiments to quantify the mechanical properties of these two DNA hydrogels in a frequency (ω) domain ranging from 1 rad s−1 to 105 rad s−1 and their viscosities (η). DLS provides an easy-to-implement experimental set-up in which only the size and concentration of the colloidal tracers has to be carefully selected. PTM provides measurements of the sample η by tracking over long times, thus significantly extending the DLS range, the relative position between two colloids embedded in the same field of view. These two methods result in valuable information about the mechanical properties of these materials and about their flow and deformation properties as functions of ω at different T.
Samples at tetrafunctional unit concentrations of 248 μM (the NS system) and 142 μM (the RG system) in H2O 100 mM NaCl buffer have been prepared. To grasp the meaning of these concentrations, these values should be compared with the values required to form a fully bonded network, respectively of ∼220 μM (the NS system30) and ∼110 μM (the RG system28). Thus, both systems lie well within the equilibrium gel31 concentration window. Samples were slowly annealed at high T to guarantee the proper self-assembly of the tetrafunctional units before the formation of the gel and sealed in 2.4 mm inner diameter borosilicate glass capillaries. Detailed observation with a video camera equipped with a 100× magnification lens confirmed that phase separation occurred for none of the investigated samples. One sample was prepared for each of the systems to evaluate their dynamic transition throughout the selected range of T.
We used a custom-built microscope equipped with a 100× Nikon objective (1.4 numerical aperture). For each T, bright field images were acquired over a temporal window of 10 min at frame-rates ranging from 10 to 50 fps. In the bright field, a 530 nm PS microsphere gives a strong intensity peak in the image when the focal plane is positioned slightly below it. We isolate the intensity peak applying a threshold to the image. Subsequently the 2D coordinates of the microsphere can be extracted by applying a “center of mass” tracking algorithm.34,35 Due to the high viscosity of the medium, the microspheres' displacements were limited (from ∼1 μm at the highest T, down to ∼20 nm at the lowest T). In order to reach an accurate measurement of the diffusion coefficient it was crucial to eliminate inevitable drifts and vibrations of the setup. These were ruled out by considering the relative position between two microspheres that were found in the same field of view. At all the investigated T, the same two particles were tracked, and the relative position vector coordinates between the two microspheres undergo a diffusive motion with an effective diffusion coefficient given by the sum of the single particles' diffusion coefficients. The diffusion coefficient was extracted from a linear fit of the measured mean square displacement (MSD), and subsequently the viscosity was calculated by means of the Stokes–Einstein relation (eqn (S17) of the ESI†). Hydrodynamic interactions between the two microspheres were neglected since the correction to the diffusion coefficient for the selected particle sizes and distances (>7 μm) is expected to be below 1%.36 Such a correction is smaller than the error introduced by the particles' radial polydispersity which is of the order of ∼5%.
![]() | ||
Fig. 2 (a) Autocorrelation functions measuring the motion of a probe particle immersed in a NS solution at a concentration of 248 μM. Experimental data is represented by circles. Solid lines indicate the result of applying CONTIN to the experimental g1(t). (b) Corresponding distribution of relaxation times as calculated from CONTIN. (c) Corresponding MSD of the probe particles (symbols). The coloured shadow backgrounds represent the statistical error calculated following Section V of the ESI.† The errors are not shown for values ΔMSD(t)/MSD(t) > 1. The figure also shows for comparison the MSD of the same tracers dispersed in a buffer of H2O and 100 mM NaCl (see Section I of the ESI†). |
Fig. 3 shows the linear viscoelastic moduli at four representative T, calculated according to the procedure described in the Methods section. At the highest T (58 °C) the viscous component G′′(ω) of the complex modulus G*(ω) takes higher values than those of the elastic modulus G′(ω) showing the typical ω and ω2 scaling at low frequency and growth consistent with a power-law with exponent 0.75 ± 0.03 at large ω. Such a power-law exponent is close to the value predicted for polymers by the Zimm model (in Θ solvent) ω2/3.39 In fact, at these T all NSs are independently floating in the solution and thus the viscous behavior of the material is expected to be predominant. On cooling (see 47 °C and 52 °C), G′(ω) and G′′(ω) start to develop more complex functional dependence. At 47 °C both moduli are characterised by parallel power-law growth ∼ωn at large frequencies with n ≈ 0.47, signatures of the proximity to a percolation point.
![]() | ||
Fig. 3 Loss (G′′) and storage (G′) moduli of the NS system for (a) 58 °C, (b) 52 °C, (c) 47 °C, and (d) 37 °C. The coloured shadows indicate the error bars calculated according to Section V in ESI.† Error bars are displayed only for ΔG′/G′ < 1 and ΔG′′/G′′ < 1. |
While the power-law dependence is a clear indication of a system approaching the percolation threshold, the transformation of n into a structural quantity such as the fractal dimension df requires some approximations. In previous work,40 the value of n has been found to vary significantly depending on the network properties, clustering around the value n = 0.5.40–43 Factors like stoichiometry, chain length and dilution of the critical gels play a role in the composition and structure of the systems and therefore significant variations in the n values have been reported.41 To name a few, in the case in which excluded volume interactions can be neglected and hydrodynamic interactions are fully screened, theory predicts42–44 that G′(ω) = G′′(ω) ∼ ω2/(df+2). For the three-dimensional percolation universality class, df = 2.5 and G′(ω) = G′′(ω) ∼ ω0.4. We observe a similarity in the response of the two moduli over a limited frequency range. Different from chemical gels where the self-similar response of moduli occurs over a wide frequency range and terminal flow is not observed, here the finite lifetime of the bonds between NSs introduces a lower-frequency crossover as observed in Fig. 3(c).
At 37 °C (Fig. 3(d)) the system is well within percolation. G′(ω) becomes dominant over G′′(ω) while G′′(ω) shows a minimum at ω ≈ 20 rad s−1, suggesting that the diffusive motion of the colloids is arrested on the corresponding timescale. In addition, G′(ω) and G′′(ω) cross over at ωc ∼ 1 rad s−1. This crossover frequency reflects the inverse of the relaxation time of the system and it conventionally indicates the transition from a fluid-like (viscous) to a rubbery, solid-like (elastic) frequency region. For comparison, at 47 °C, the crossover is found at a higher ω, in accordance with a faster relaxation time of the probe colloid (see Fig. 2(a)) for a weaker network stabilised at higher T. In all the investigated samples, at ω < ωc terminal fluid behavior is recovered with G′(ω) ∼ ω2 and G′′(ω) ∼ ω1.
When a clear plateau in G′(ω) is present, its value can be used to provide an estimate of the cage size ξ confining the probe particle (of radius a), according to the relation in eqn (S15) of the ESI.† We observe plateau values of G′(ω) up to 7 Pa, corresponding to ξ ≈ 26 nm. As a reference, we note that the order of magnitude of the elastic modulus can be estimated by the product of the density of elastically effective chains times the thermal energy.45,46 In the fully bonded low T limit, each NS contributes to two effective chains, suggesting as the (DLS experimentally unaccessible) low T limiting value G′(ωplateau) ≈ 600 Pa. Unfortunately, as commented before, at such low T g1(t) would decorrelate at much longer times than the maximum accessible experimental time of DLS techniques.
The corresponding experimental g1(t) for the RG system are displayed in Fig. 4(a) together with their CONTIN analysis. The g1(t) functions clearly show an increase in the relaxation times on cooling from 50 °C to 32 °C. On further lowering T the relaxation times become faster and faster, recovering a typical fluid-like value at the lowest investigated T. This behavior is replicated in the associated MSD curves in Fig. 4(c) for the high T regime (where the dynamics gets slower on cooling) and in Fig. 4(d) for low T (where the dynamics gets faster on cooling). Thus, the viscous high and low T behavior is comprised (see the MSD at T ≈ 30 °C) of the caging regime typical of non linear viscoelastic systems, similar to the one observed in the NS system at low T (see Fig. 2(c)).
![]() | ||
Fig. 4 (a) g1(t) measuring the motion of a probe particle dissolved in the RG hydrogel at a NS concentration of 110 μM. Experimental data is represented by circles and the result of applying CONTIN procedures is indicated with solid lines. (b) Corresponding distribution of relaxation times as calculated from CONTIN. (c) MSD for T ranging between 50 °C and 32 °C and (d) for T ranging between 28 °C and 5 °C. The coloured shadow backgrounds represent the statistical error calculated following Section V of the ESI.† The errors are not shown for values ΔMSD(t) > MSD(t). |
Fig. 5 shows the corresponding viscoelastic moduli for different characteristic T. The data clearly show on cooling the viscoelastic behavior typical of a fluid (a), of a percolating system (b), of an elastic gel (c), again of a percolating system (d) and of a low T fluid (e). Indeed, panels (a) and (e) are characterised by G′′(ω) > G′(ω) in such a way that the loss modulus prevails over the elastic modulus in the absence of crossovers in the whole ω range. Panels (b) and (d) are characterised by G′′(ω) ≈ G′(ω) for ω > 102 rad s−1 and by clear power-law frequency dependences with a scaling exponent n again around 0.5 (best fit values of n = 0.46 and n = 0.58 respectively). Finally, panel (c) shows the typical behavior of a viscoelastic solid where G′′(ω) shows a minimum at ω ≈ 500 rad s−1 and a corresponding plateau value of G′(ω) of about 8 Pa. We associate this behavior with an elastic response where the colloidal particle is trapped in a cage. It must be noted that our DNA hydrogels are physical gels in which the bond breaking and formation events, absent in chemical gels, are responsible for the final relaxation behaviour observed in the low frequency region. This inevitable crossover to standard low frequency relaxation determines the observed crossover between G′(ω) and G′′(ω) at 28 °C.
![]() | ||
Fig. 5 G′ and G′′ moduli of the RG system from high (top) to low (bottom) T, showing mechanical features characteristic of different states: (a) a fluid at 50 °C, (b) close to the percolation threshold at 36 °C, (c) a viscoelastic solid at 28 °C, again (d) close to percolation at 20 °C and (e) a fluid at 5 °C. The coloured shadows indicate the error bars calculated according to Section V in ESI.† |
![]() | (1) |
![]() | ||
Fig. 6 Comparison of the viscosities η in the RG (light and dark blue, and turquoise symbols) and in the NS (red, dark and light orange, and yellow symbols) systems. The inset shows the Arrhenius plot for the NS system in the gel region (at T below the melting temperature). The dashed line represents the best fitting with an exponential Arrhenius-like function with an activation energy EA ≈ 105 kcal mol−1. The error bars are calculated as explained in Sections III and V of the ESI.† |
In addition, it is possible to estimate the bulk viscositiy η directly from the rheological data as the limit for ω → 0 of G′′(ω)/ω. As shown in Fig. 6, the reported values by the three methods are all consistent.
Beside providing a set of independent measurements to validate the η values estimated from DLS, the PTM technique gives access to a different time window compared to DLS and allows us to extend the range over which η measurements are possible by a further three orders of magnitude.
From the PTM data we extract the positions over time of two tracer particles to compute the interparticle displacement, which undergoes a diffusive motion that is not affected by systematic errors associated with any source of global drift in the optical set-up. As discussed in detail in Section III of the ESI,† the selected exposure time does not allow us to quantify the amplitude of the short-time vibrational motion of the probe that would result in a positive t = 0 intercept of the MSD. In addition, the finite exposure time, as well as the tracking precision, introduces artefacts that make unphysical any interpretation of the intercept values.47 These values have been subtracted in the MSD presented in Fig. 7. Nevertheless, all this does not affect the evaluation of the slope of the MSD vs. time and thus of the diffusion coefficient, allowing us to compute η via the Stokes–Einstein relation.
![]() | ||
Fig. 7 MSD measured via the PTM technique at T ranging from 21.4 °C up to 38.6 °C. The time-independent offset, which is affected by the finite exposure time and by tracking artefacts, has been subtracted from the curves. The error bars are computed as detailed in Section III of the ESI.† |
Around T ≈ 38 °C, where both DLS and PTM measurements are possible, the agreement between DLS and PTM is highly satisfactory. On further cooling, η progressively increases as a result of the longer lifetime of the interparticle bonds. As expected for a strong network-forming fluid,48 at low T, η follows an Arrhenius behavior with an activation energy EA ≈ 105 kcal mol−1 which reflects the enthalpy of the inter particle bonds (44.6 kcal mol−1 (ref. 49)).
For the RG, η was evaluated via DLS, also shown in Fig. 6. The T-dependence of η shows a clear maximum at intermediate T. Consistently with the behavior displayed by the linear viscoelastic moduli, η first increases on cooling when the network progressively forms and then decreases again when the competitor sequences start binding, capping the NS sticky ends until the point at which, at very low T, the network is completely disentangled and fluid-like behavior is recovered. At this point a η comparable to the one reported at 50 °C is again observed, which is only few times higher than the solvent η. In the case of the RG, an accurate estimate of the activation energy is not possible, since the system is designed to evolve back to a fluid at low T. In the region of T where η increases, the slope of lnη vs. T−1 progressively changes with T preventing an accurate estimate of the activation energy.
The frequency-dependent viscoelastic moduli measured with DLS passive microrheology revealed the different viscoelastic properties of the two systems as a function of T. The data showed the crossover from a fluid to a gel for the NS system via a percolation transition, and a multiple sequence of transitions from a fluid to a percolating system to an elastic gel to a percolating system and to a fluid again on cooling, for the RG system. Similarly, η monotonically increased on cooling in the NS system while it first increased and then decreased again in the RG system. The combination of DLS and PTM techniques allowed us to explore more than six orders of magnitude in η.
In conclusion, we provide evidence of the possibility of measuring material properties in hydrogels built by controlled association of DNA-made particles, with specified functionality and binding sequences. The comparison between the NS and RG systems highlights both the flexibility offered by the design of the DNA sequences and the possibility to tune the mechanical response of these biocompatible all-DNA materials with minor changes in the DNA sequences.53 The biocompatibility and thermoreversible character of these materials opens the possibility for a large number of applications related to the biomedicine and drug delivery fields in which fine control of the materials' mechanical properties is required.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c8sm00751a |
This journal is © The Royal Society of Chemistry 2018 |