Introduction

The interaction between pathogens and host proteins play an important role in the progression of various infections. Fibrin, one of the most abundant plasma proteins, is most widely recognized for its key role in haemostasis1,2. The coagulation cascade facilitates timely formation and dissolution of fibrin clots for bleeding control and wound healing response. However, fibrin also plays a major role in the host defence against invading pathogens1,2. The fibrin scaffolds, which can be rapidly formed and dissolved, can form a physical barrier that blocks invasion of pathogen, entrap pathogens within fibrous mesh to prevent dissemination, as well as help direct immune cell function2,3. Due to its crucial role in innate immunity, fibrin (and the coagulation cascade) is also a major target for manipulation by several bacterial species to establish host infection4,5. For instance, these bacteria can hijack the fibrin scaffolds to form aggregates or colonize the surface of medical implants, hiding within an insoluble film that protects them against both immune cells and antibiotic drugs while the bacterial population grows6,7,8. After reaching a critical population, certain bacteria can also trigger fibrinolysis to dissolve the surrounding fibrin mesh and systemically disseminate9,10. A prime example is Staphylococcus aureus, a common non-motile bacteria and a leading cause of hospital-acquired infections11, which develops persistent infection and antibiotic resistance through its manipulation of host fibrin5. Thus, bacteria-fibrin interaction is at the crux of infection and has prompted not only antibacterial, but also antibiofilm strategies to combat bacterial persistence and antibiotic resistance10,12,13,14,15.

The interaction of bacteria with host fibrin is dynamic and evolves through a series of orchestrated fibrin binding, fibrin clot formation (coagulation), and fibrin clot dissolution (fibrinolysis) events. At the nano- to microscale, the bacteria-fibrin interaction is underpinned by the dynamical behaviour of the bacterial cells and the innate microstructural dynamics of the extracellular network scaffolds16, both of which span a broad range of timescales down to the sub-milliseconds. A suite of techniques has enabled the biophysical study of bacteria interaction with fibrin, from direct counting to determine the number of bacteria in fibrin scaffolds17 to advanced microscopy techniques such as scanning electron microscopy (SEM)8 and confocal fluorescence microscopy18 to visualize the morphology of bacteria-infected clots. A combination of turbidimetry, SEM, X-ray scattering, and mechanical rheometry has been utilised to investigate the ultrastructure, lytic stability, and viscoelastic properties of fibrin clots formed by staphyloccocal products19. However, these conventional techniques only provide a snapshot of information that cannot capture the complex bacteria-fibrin dynamics spanning multiple timescales. Understanding these multiscale dynamics over the course of the formation and dissolution of fibrin scaffolds may provide unique insights on the pathogenesis of bacterial infection and progression to lethal sepsis, beyond static information available from conventional techniques.

We present a non-invasive optical approach based on laser speckle fluctuation to investigate, in real-time, the multiscale dynamics of bacteria during fibrin clot formation and fibrinolytic dissolution, using a model system of polystyrene microspheres in fibrin scaffolds. Inert microspheres have been utilized as a model to help elucidate the biophysical aspects of bacterial behaviour inside the extracellular matrix, independent of the milieu of biochemical and genetic factors. Dunn et al. employed polystyrene microspheres to show that S. aureus and Escherichia coli were incorporated into polymerizing fibrin scaffolds in the same numbers as inert particles of similar size even in the absence of bacteria-specific surface proteins17. Using different levels of bovine serum albumin coating on polystyrene microspheres, Ma et al. further showed that the impact of Staphylococcus epidermidis on the microstructure of fibrin clots was mediated by the adherence of fibrin to the surface of bacterial cells18. Our approach harnesses speckles that arise from light scattered by the particles (our model for spherical, non-motile bacteria such as S. aureus and S. epidermidis) to directly analyse their nano- to microscale motions within the fibrin scaffolds. Coupled with high-speed acquisition, the timescale-dependent analysis of speckle fluctuation provides information on the dynamical behaviours of the particles across a broad cascade of timescales from sub-milliseconds to seconds. Unlike conventional particle tracking approaches14,18,20,21,22,23,24,25, the use of laser speckle circumvents the need for an elaborate high-resolution microscope to precisely visualize and track individual particles, enabling rapid real-time monitoring of particle dynamics in situ.

We first discuss the particle dynamics and biophysical insights that can be gleaned from the timescale-dependent analysis of laser speckle fluctuation. Then, we demonstrate the utility of our approach by investigating the dynamics of particles with different sizes and surface properties as they become enmeshed in polymerizing fibrin scaffolds. Finally, we apply our approach to monitor particle dynamics in real-time during thrombin-mediated plasma clot formation and plasminogen-mediated fibrinolysis in human plasma samples. Our approach can be readily applied to study various in vitro and in vivo grown bacteria-infected clots and other bacterial biofilms in the future.

Results

Characterizing multiscale dynamics of microparticles in extracellular network with laser speckle fluctuation

We utilize inert polystyrene microparticles to physically model spherical non-motile bacterial cells17,18, as a means to emulate bacterial dynamics within an extracellular network scaffold. In particular, we leverage dynamic laser speckle patterns to quantify and characterize the nano- and micro-scale motions of particles via their mean square displacements (MSD) within fibrin scaffolds during coagulation and fibrinolysis in situ.

Coherent laser light scattered by the particles interferes to form a granular pattern—speckle—that temporally fluctuates at the rate governed by particle dynamics (Fig. 1a). We utilize a high-speed camera to record a time series of fluctuating speckle patterns at sub-MHz frame rate in order to capture rapid particle dynamics down to the microsecond timescale (see Methods for details of the optical setup). Unlike traditional particle tracking methods, it is not necessary to track the precise position of individual particles over time with a high-resolution microscope or advanced position detection methods to obtain the particle MSD. Instead, the rate of speckle intensity fluctuation is quantified by the ensemble-average speckle intensity autocorrelation function, g2(τ), where τ denotes the decorrelation timescale (Eq. (1) in Methods). Using an empirical approximation of the diffusing wave spectroscopy formulation (Eq. (2) in Methods)26,27,28,29, the intensity autocorrelation function provides an estimate of the particle MSD, 〈Δr2(τ)〉 (Fig. 1b, top). This approach allows minute displacements of the particles to be detected with sub-nanometre sensitivity over a broad range of timescales (Supplementary Fig. 1)29,30. The MSD exhibits timescale-dependent behaviours that can be described via local power scaling law, α(τ), where 〈Δr2(τ)〉∝τα(τ) and α(τ) is obtained from the log-log derivative of the MSD w.r.t τ (Eq. (3) in Methods).

Fig. 1: Characterization of particle dynamics in extracellular networks.
figure 1

a Schematic of laser speckle fluctuation measurement in fibrin scaffolds with incorporated microparticles. Upon laser illumination, light scattered by particles interferes to form a speckle pattern captured by a high-speed camera. Motions of the particles within the sample produce fluctuation in the speckle intensity, which can be analysed via the characteristic intensity autocorrelation, g2, as a function of lag time, τ, to obtain the particle MSD. b Examples of timescale-dependent particle MSD, 〈Δr2(τ)〉, and its local power scaling exponent, α(τ), measured in polymerized fibrin gel (dark blue) and unpolymerized aqueous solution of fibrinogen (light blue). Insets illustrate representative behaviour of a particle in the extracellular network over short- (Regime I), intermediate- (Regime II), and long- (Regime III) timescales.

Full size image

Although the MSD by itself indicates the extent of the motions undertaken by the particles at any given timescale, it is the timescale-dependent behaviour described by the power law α that provides insights on the biophysical nature governing the dynamics of the particles. For instance, α = 1 signifies freely diffusive Brownian (thermal) motion in linear viscous fluid, where MSD scales linearly with time, as is seen with particles in an aqueous solution of fibrinogen molecules (Fig. 1b, light blue curve). Beyond this limit, α > 1 implies particle dynamics is super-diffusive, which can be a result of actively driven (e.g., swimming bacteria20, self-motile particles31, molecular motor32, cell forces30) or directional (e.g., inside nanotubes33) motion. At the other limit, α = 0 indicates motion is completely restricted, either by geometric (e.g., fibrous network structure) or elastic (linear elastic solid) ‘cage’. Between the two limits, 0 < α < 1 characterizes a range of sub-diffusive dynamics that can arise from several factors, including steric hindrance34,35, adhesion or binding of particles to the network structure36, macromolecular crowding37, and linear viscoelasticity29,30.

Unlike the simple diffusive dynamics of particles in fibrinogen solution where α = 1, particles incorporated into a fully polymerized fibrin hydrogel scaffold exhibit a range of timescale-dependent α values that span sub-diffusive to super-diffusive dynamics (Fig. 1b, dark blue curve). This timescale-dependent dynamics may be interpreted in three distinct temporal regimes corresponding to the behaviours of a small particle—our model for a spherical bacterial cell—traversing a fibrous extracellular network at different timescales (Fig. 1b, cartoon illustration). At the shortest timescale (Regime I), the particle explores the fluid-filled space via short excursions within its local environment of the fibrous network. The particle dynamics may be sub-diffusive to diffusive, where the short-time power scaling law, αI, is governed by factors such as the local viscoelasticity of the microenvironment29,30, particle-to-mesh size ratio (steric hindrance)34,35 and binding of particles to the network structure36. After some time (Regime II), the particle traverses far enough to reach the fibrous structures forming its immediate ‘cage’ that constrains further displacement. The particle dynamics becomes almost completely restricted; MSD exhibits a plateau with value 〈Δr2〉0II, corresponding to a minimum α value, α0II, that approaches 0. At even longer timescale (Regime III), the particle may ‘escape’ from its local environment following interactions with its immediate fibrous structures, aided by the total relaxation of the network16,38. At a certain transition timescale, τT, the particle dynamics becomes more diffusive than restricted with α(τ) exceeding 0.5. In this example, α(τ) continues to rise above 1, entering the super-diffusive regime which may be driven by a combination of the residual stress in the freshly polymerized fibrin network38, stress driven by the localized swelling/shrinking within the fibrin network due to syneresis39,40, and the momentum transfer as the particles collide with these elastic fibrin fibres. The overall long-time power scaling law, αIII, reflects extent to which the particle is freed from its immediate ‘cage’ within the fibrous scaffold.

Thus, analysis of laser speckle fluctuation provides a non-invasive tool to extract multiscale dynamics parameters of particles such as bacterial cells enmeshed within extracellular network scaffolds in situ. These parameters offer insights on the biophysical interactions of particles with the extracellular network, including factors such as particle-to-mesh size ratio, particle-network binding, and network dynamics, beyond the static structural information provided by conventional microscopy. In the following sections, we demonstrate the applications of our approach to real-time in situ monitoring of 1) the incorporation of particles of varying size and surface chemistry into polymerizing fibrin scaffolds, and 2) trapping and release of particles within fibrin scaffolds during human plasma clot formation and fibrinolytic dissolution.

Real-time monitoring of particle dynamics enmeshed in polymerizing fibrin scaffolds

An important mechanism that facilitates bacterial survival and antibiotic resistance in the human body is the interaction of bacteria with host fibrin to form protective fibrin scaffolds6,7,8. To demonstrate the applicability of our laser speckle fluctuation approach for investigating this important process, we use a model system of polystyrene microspheres of sizes 0.43, 0.70, and 2.65 μm in diameter (measured with Malvern, Zetasizer) incorporated into polymerizing fibrin hydrogel scaffolds. These particle sizes were chosen to emulate the dimensions of single staphylococcal bacteria and bacterial aggregates11. Partially cross-linked fibrin scaffolds were prepared with fluorescence-labelled plasminogen-depleted human fibrinogen and human α-thrombin in HBS buffer (see Methods for sample preparation). Due to binding affinity of fibrinogen protein to polystyrene, the particles were surface-modified with polyethylene glycol (PEG) coating to prevent severe clumping of particles that drastically disrupted the fibrin network structure36 (Supplementary Fig. 2a).

The microstructure of polymerizing fibrin network and fully polymerized fibrin scaffolds was visualized in real time during polymerization with confocal fluorescence microscopy. Fibrin fibre segments were visible within polymerization time of Δt = 0.5 min and most soluble fibrinogen molecules (background fluorescence) were formed into the fibrin network in under 10 min (Fig. 2a, top). The median mesh size measured from confocal fluorescence microscopy images of the fibrin network decreases from ~3 μm at Δt = 1 min to ~2.2 μm at Δt > 25 min (Fig. 2c). Notably, the incorporation of polystyrene beads did not alter the mesh size of the fully polymerized fibrin scaffolds; there was no significant difference in the mesh size distribution measured at Δt = 60 min across all four samples (Fig. 2g). However, the confocal fluorescence images reveal distinct ways in which beads of different sizes are incorporated into the fibrin network (Fig. 2a, bottom), owing to the different capacity for fibrin(ogen) to adhere to the surface of each particle size following PEG coating (Supplementary Fig. 2b). (This difference likely stems from the size-dependent efficiency of the PEGylation procedure, which is expected to be most effective for particles ≥1 μm in diameter36,41.) The 0.43 and 0.70-μm diameter beads appear with strong fluorescence signal in small aggregates, indicating that they are coated by immobilized fibrinogen molecules (fluorescence-labelled) (Supplementary Fig. 2c). As a result, these particles generally form the nodes of the fibrin meshwork with fibres branching out or attach to the fibre structures (white arrows in Fig. 2a). These fibrin(ogen)-bound beads may mimic the behaviour of binding proteins on the bacterial cell wall (e.g., ClfA on S. aureus), which binds soluble and immobilized fibrinogen to form bacterial aggregates42,43. In contrast, the 2.65-μm diameter beads are not visible on the fluorescence channel, indicating sufficient PEG coating to prevent attachment of fibrin(ogen). Brightfield image shows that these beads occupy the void spaces of the fibrin network (Supplementary Fig. 2d).

Fig. 2: Dynamics of beads in polymerizing fibrin scaffolds.
figure 2

a Confocal fluorescence images of pure fibrin gel at different polymerization time, Δt, (top, single plane) and fully polymerized fibrin gels containing no bead, 0.43-, 0.70-, and 2.65-μm diameter polystyrene beads (bottom, 5-μm Z-stack). (Images of the early polymerization process with beads are provided in Supplementary Fig. 2d.) Arrows point to fibrinogen-coated beads at the nodes of the fibrin network or attached to fibrin fibres. MIP: maximum intensity projection. Scale bar: 10 μm. b Time-lapsed measurement of timescale-dependent particle MSD and α of 0.43-μm beads over the course of polymerization (blue: early time, red: fully polymerized). c–f Fibrin mesh size and particle dynamics parameters: αI, α0II, and αIII, extracted from the measured timescale-dependent MSD and α as a function of polymerization time Δt for 0.43- (blue), and 0.70- (green), and 2.65-μm (red) beads. For αI and αIII, line and shaded area represent slope and 95% confidence interval of linear regression, respectively. g–j Repeated measurements of mesh size, αI, α0II, and αIII in fully polymerized fibrin gels at Δt = 24 h (N = 9 from 3 measurement locations in each of 3 biological replicates). Violin shape, white marker, and solid bar represent the kernel distribution, median, and interquartile range, respectively. In (j), measurement immediately after full polymerization at Δt = 60 min is also shown with (×). Z and P indicate normalized test statistics and P-value of Cuzick’s test for trend across ordered groups. Symbols indicate levels of statistically significant difference per Kruskal–Wallis test when compared to no bead (*), 0.43- (†) and 0.70-μm (‡).

Full size image

The dynamics of particles interacting with polymerizing fibrin scaffolds is monitored in real-time with laser speckle fluctuation. Timescale-dependent MSD and α(τ) were measured between Δt = 0.5–60 min, with the shortest time interval of ~22 s to capture the initial rapid formation of fibrin fibre structures (Fig. 2b). The particle MSD exhibits the general pattern described above of power scaling at the shortest timescale (τ < 10–2 s), followed by plateau behaviour at intermediate timescale (10–2 < τ < 10–1 s), and once again power scaling at long timescale (τ > 10–1 s) throughout the polymerization process. Multiscale particle dynamics parameters, αI, α0II, and αIII, were extracted over the course of fibrin polymerization for each of the three bead sizes (Fig. 2c–f). Statistical comparison of the parameters between fibrin scaffolds with no bead and different bead sizes are also investigated for the fully polymerized clots after Δt = 24 h (Fig. 2g–j).

In Regime I, αI of the 0.43-μm beads (Fig. 2d, blue) is nearly 1 at Δt = 1 min, indicating almost free diffusion of the beads and fibrin(ogen) complex that rapidly formed at the initial phase of polymerization (Supplementary Fig. 2d). As the mesh size decreases over the course of polymerization, the increasing steric hindrance results in progressively more sub-diffusive dynamics, corresponding to a decreasing trend in αI as a function of Δt. Notably, the decrease in αI exhibits a linear trend as a function of the inverse of the fibrin mesh size (Supplementary Fig. 3a), consistent with previous observation of anomalous sub-diffusion in F-actin network34. Similar trend is observed for the 0.7-μm beads (Fig. 2d, green), which are also bound to fibrin(ogen), but with overall lower αI values compared to the 0.43-μm beads due to the larger particle-to-mesh size ratio causing greater steric hindrance. Distinctly, the 2.65-μm beads (Fig. 2d, red), which are not fibrin(ogen)-bound, exhibit the highest αI during the early phase of polymerization even with the largest bead size. However, as the mesh size becomes smaller than the bead size at Δt ~ 4 min, αI decreases below 0.5 indicating particle dynamics that is closer to being trapped in the fibrin mesh than being able to freely diffuse in the void spaces. In the fully polymerized fibrin clots, there is a decreasing trend in αI with increasing bead size (Fig. 2h, P = 4.8 × 10–7), likely governed by the steric hindrance of larger particle-to-mesh size ratio.

In Regime II, the fibrin(ogen)-bound 0.43- and 0.70-μm beads exhibit plateau behaviour with α0II ~ 0.1 throughout the polymerization process (Fig. 2e, blue and green), indicating relatively restricted motion within the forming fibrin meshwork. The largest α0II value is observed with the non-adherent 2.65-μm beads at early phase of polymerization (Fig. 2e, red), before the beads become increasingly trapped by the smaller mesh size of the fibrin network at Δt > 4 min. In the fully polymerized clots, the α0II value is governed by the relative viscoelasticity29,30 of the fibrin network as experienced at different length scales of the three bead sizes. Thus, there is a decreasing trend in α0II with bead size (Fig. 2i, P = 1.0 × 10–4) as the large 2.65-μm beads probe the collective network elasticity of the fibrin scaffold whereas the smaller beads experience more viscous contribution from the fluid microenvironment in the void spaces. Meanwhile, the MSD value at the plateau, 〈Δr2〉0II, of all fully polymerized samples (all three bead sizes and scaffold with no bead) converges to the same value after scaling by the particle sizes (Supplementary Fig. 3b). This confirms that all four fibrin scaffolds exhibit the same shear modulus magnitude according to the Generalized Stokes-Einstein Relation44,45,46 (see equation in Supplementary Fig. 3b) as they were prepared under the same condition other than the incorporation of different bead sizes.

In Regime III, αIII of all bead sizes exceeds 1 and increases as a function of Δt (Fig. 2f), indicating increasing contribution of active (super-diffusive) particle dynamics as the polymerization progresses. Similar behaviour has previously been observed during the polymerization of polyacrylamide hydrogels, where the source of the active dynamics was attributed to the stress incurred within the network structures as a result of forming connected fibre cross-links38. The increasing trend in αIII as a function of Δt in Fig. 2f is in agreement with this interpretation as the formation of more cross-links over time would incur greater stress in the fibrin network. Furthermore, higher values of αIII are observed at Δt = 60 min (Fig. 2j, ×), immediately after reaching full polymerization, compared to at 24 h (Fig. 2j, violin), after the stress incurred from polymerization has gradually relaxed over time38. In addition, the timescale at which active dynamics become dominant (i.e., τ at which α(τ)>1) is approximately the same for all bead sizes throughout the polymerization process (Supplementary Fig. 3c), further supporting our framework that the observed super-diffusion partly stems from the polymerization of fibrin itself. Another phenomenon that could contribute to the observed super-diffusive dynamics in the fully polymerized fibrin scaffolds is syneresis, which can cause local shrinking and swelling of the polymer network that, due to the elasticity of the fibre network, incurs localized stress39,40. In colloidal gels, anomalous super-diffusive behaviour with α = 1.5 has been observed at long timescales and associated with the effect of syneresis39.

Whereas these two mechanisms discussed above (polymerization-incurred stress and syneresis) are governed by the fibrin network itself, an increasing trend in αIII is observed as a function of increasing bead size (Fig. 2j, P = 6.5 × 10–4), indicating that the particles also participate in driving the extent of the observed super-diffusive dynamics. We postulate that, when a diffusing particle interacts with the elastic fibrin network (the local shrinking/swelling of the network due to syneresis could further drive these interactions), a transfer of linear momentum occurs that amplifies the active dynamics of the beads, resulting in super-diffusive behaviour. Particularly, the α behaviour does not stay constant in Regime III; rather, α(τ) rises to a super-diffusive maximum before trending back down toward a normal diffusive behaviour (Fig. 2b, bottom). This peak may reflect the timescale of the particle-network interaction that rapidly increases α as the momentum is transferred to the particle. Then, the following decrease of α may suggest that the particle either return into the pore space or escape the local mesh after its interactions with the fibrin network, allowing it to return to a more diffusive behaviour. Furthermore, the increasing trend in αIII as a function of polymerization time (Fig. 2f) is consistent with the increasing network stress, which would enhance the momentum transferred to the particle during these particle-network interactions, as more crosslinks are being formed over the course of the polymerization process.

Our results show that microparticles exhibit multiscale dynamics within the fibrin scaffolds over the course of fibrin polymerization. Multiscale parameters αI, α0II, and αIII, obtained from the timescale-dependent analysis of laser speckle fluctuation reveal distinct dynamics of particles with different size and surface chemistry, emulating different characteristics of bacterial cells. The distinct multiscale dynamical behaviours demonstrated here cannot be perceived from structural visualization of the particles and the fibrin network structure alone, and thus, may provide unique insights on the process of fibrin clot formation during bacterial infection.

Trapping and release of particles during plasma clot formation and fibrinolytic dissolution

Certain bacterial species such as S. aureus not only promotes fibrin clot formation via its coagulases, but also secretes staphylokinase to activates host plasminogen and induce clot dissolution via fibrinolysis9,10. The picture that has emerged is that fibrin(ogen) binding and early release of coagulases help form protective fibrin scaffolds around proliferating bacteria, while later release of staphylokinase allows the bacterial population to disperse and cause systemic infection2,5. On the other hand, fibrinolysis also serves as a potential mechanism for antibiofilm therapies to combat bacterial persistence and antibiotic resistance10,12,14. To demonstrate the application of our laser speckle fluctuation approach for investigating bacterial dynamics during the entire coagulation-fibrinolysis cascade, we monitor in real-time the dynamics of bead entrapment during plasma clot formation followed by bead dispersion during fibrinolytic clot dissolution. We obtained platelet poor plasma from clinical whole blood samples according to the MGH Core Laboratory procedure (MGH IRB#2017P000419). The plasma sample was spiked with clinical thrombin reagent and tissue type plasminogen activator (tPA) to initiate clotting and fibrinolysis, respectively (see Methods for sample preparation). We chose tPA in this study for its clinical relevance as an investigative strategy to combat bacterial biofilms10,12,13.

The microstructure of the plasma clot during coagulation and fibrinolysis was monitored with time-lapsed reflectance confocal microscopy (Fig. 3a), enabled by spectrally encoded confocal microscopy (SECM)47. (Unlike the purified fibrin scaffolds in Fig. 2, fibrinogen in the clinical plasma sample was not fluorescence-labelled.) Measurement of fibre density from the SECM images48 shows that maximum clot formation is achieved at coagulation time Δt = 15 min, followed by a brief latent period before clot dissolution is visibly observed after Δt > 20 min (Fig. 3b, dots). Meanwhile, mesh size of the plasma clot remains relatively unchanged at approximately 2–3 μm up to Δt = 35 min (corresponding to ~50% of peak fibre density) before rapidly expanding beyond this point (Fig. 3b, line). This observation is consistent with recent findings of pore expansion that drives internal fibrinolysis in human plasma clots49.

Fig. 3: Monitoring particle dynamics during plasma clot formation and fibrinolytic dissolution.
figure 3

a Confocal reflectance images of plasma clot at different time points during coagulation and fibrinolysis. Scale bar: 20 μm. b Fibrin fibre density (dots) and mesh size (line) measured from the images in (a) as a function of coagulation time, Δt. Line and shaded area represents mean ± s.d. of mesh size distribution across a field-of-view of 1.8 mm × 0.1 mm. c Time-lapsed measurement of timescale-dependent particle MSD and α over the course of polymerization and fibrinolysis (blue: early polymerization, red: after lysis) in plasma clot with fibrin-bound uncoated beads. d Time-lapsed measurement of particle MSD during clotting and lysis of clot from the same plasma sample as in (c), but with free (i.e., do not adhere to fibrin) PEG-coated beads. e, f Confocal reflectance images of two fully polymerized clots (i.e., tPA was not added) from the same plasma sample with uncoated and PEG-coated beads, respectively. Scale bar: 20 μm (large image) and 5 μm (coloured box). g–j Particle dynamics parameters: αI, α0II, αIII, and τT, extracted from the measured timescale-dependent MSD and α as a function of coagulation time Δt with bound uncoated (blue) and free PEG-coated (red) beads. Shaded area indicates time points after lysis where particle dynamics approaches diffusion in viscous fluid and distinct behaviour in Regimes II and III are not applicable. For αI and αIII, line and shaded area represent slope and 95% confidence interval of linear regression, respectively.

Full size image

To model bacterial cells expressing different levels of surface binding proteins (e.g., ClfA in S. aureus), we employed uncoated (more fibrin binding) and PEG-coated (less fibrin binding) 0.43-μm diameter polystyrene beads. On the one hand, the uncoated beads bind to fibrin(ogen) in the plasma, resulting in multiple small (Fig. 3e, light blue) and large aggregates (Fig. 3e, dark blue) trapped within the formed fibrin scaffold. Compared to the pure plasma clot in Fig. 1a, the incorporation of fibrin(ogen)-bound beads results in distinctly heterogeneous fibre network structure, not unlike that of fibrin hydrogel infected by S. epidermidis18 and plasma clot induced by wild-type S. aureus strains8. Timescale-dependent MSD and α(τ) were measured between Δt = 1 to 54 min, tracking the initial coagulation phase (Fig. 3c, blue to green hues) followed by fibrinolysis phase (Fig. 3c, yellow to red hues). Particle dynamics in the coagulation phase exhibits distinct multiscale behaviour across three regimes, similar to the behaviour during fibrin polymerization in Fig. 2b. In the fibrinolysis phase, the particle dynamics gradually approaches that of freely diffusive motion with an MSD power scaling law of α → 1 as the fibre structure dissolves away. On the other hand, the PEG-coated beads minimally bind to fibrin(ogen) and are incorporated into the void spaces in the formed fibrin scaffold, either as individual beads (Fig. 3f, orange) or small aggregates (Fig. 3f, red), with little to no attachment to the fibre structure. Unlike the uncoated beads, the measured timescale-dependent MSD of the PEG-coated beads exhibits only one distinct regime particle dynamics with MSD power scaling law of α → 1 through the entire coagulation and fibrinolysis phases (Fig. 3d). The particle dynamics closely resembles that of the freely diffusive particles in unpolymerized fibrinogen solution in Fig. 1b, corroborating the SECM images (Fig. 3f) that show the particles exploring the fluid microenvironment inside the void spaces of the clot scaffold. This behaviour notably differs from that in Fig. 2 where the PEG-coated 0.43-μm beads are bound to the fibrin network in the purified fibrin scaffolds, suggesting that fibrin(ogen) in clinical blood plasma may be less prone to surface immobilization than in the purified reconstitution.

Multiscale dynamics parameters, αI, α0II, αIII, and τT, were extracted from the timescale-dependent MSD and α(τ) to quantitatively monitor the progression of particle dynamics over the course of plasma clot formation and fibrinolytic dissolution (Fig. 3g–j). In Regime I, the short-time power scaling law, αI, of the free (PEG-coated) beads (Fig. 3g, red) varies dynamically over time, allowing the coagulation and the fibrinolysis phases to be clearly discerned. During coagulation, αI decreases sharply to a minimum value of ~0.8 at Δt = 15 to 25 min, which corresponds to the period of peak clot formation (i.e., peak fibre density). During fibrinolysis, αI increases to ~1 with the sharpest rise between Δt = 30–40 min, which corresponds to the rapid expansion of the network mesh size in Fig. 3b, after which both αI and the mesh size plateau out at their maximum values. Compared to the free beads, the bound (uncoated) beads exhibit overall lower αI values (Fig. 3g, blue), which would be expected from a more restricted particle dynamics due to binding to the fibre structure. Unlike the free beads, αI is sub-diffusive (αI ~ 0.75) from the early stage of coagulation (rather than decreasing sharply during the coagulation phase) and the increase in αI during the fibrinolysis phase is more gradual. This behaviour likely reflects the dynamics of particle-fibre complexes (due to attachment of bound beads to fibre structure), which would experience greater steric hindrance than the particles alone while the fibrin scaffold is being formed or dissolved. Notably, the progression of αI as a function of Δt for the bound beads roughly follows that of the fibre density in Fig. 3b, suggesting that the dynamics of the bound beads is tightly coupled to that of the fibre structure.

In Regime II, the minimum α value, α0II, of bound beads are close to 0 (Fig. 3h), indicating restricted particle dynamics similar to the results in purified fibrin scaffold in Fig. 2d. Nevertheless, variation in α0II is apparent through the coagulation and fibrinolysis phases. Given that beads are bound to the fibre structure, α0II value is governed by the relative viscoelasticity29,30 of the plasma clot at the length scale of bead aggregates. During the coagulation phase, α0II decreases sharply to a minimum value of ~0.04 at Δt = 10 to 20 min, which corresponds to the period of peak fibre density, indicating an increase in elasticity of the clot as more fibre structure is formed. During the fibrinolysis phase, α0II monotonically increases between Δt = 20 to 40 min, roughly tracking the decrease in fibre density as the clot elasticity decreases with the dissolving fibres.

In Regime III, the long-time power scaling law, αIII, of bound beads exhibits a wide range of dynamics from almost completely restricted motion (αIII → 0) to free diffusion (αIII → 1) over the course of coagulation and fibrinolysis (Fig. 3i). At this long timescale, αIII sharply decreases from a maximum value of ~1 to a minimum value of ~0.2 within the span of Δt = 10 min, indicating the beads initially diffuse freely in the plasma before becoming trapped within the coagulated clot. Following a brief latent period, αIII rapidly returns to its maximum value of 1 between Δt = 20 to 35 min during the fibrinolysis phase, indicating the release of the entrapped beads as fibrous scaffold dissolves away. Another dynamics parameter that may offer insights on this release from entrapment is the transition timescale, τT, where α(τ) exceeds 0.5 after rising from the minimum value α0II (see Fig. 1b). Here, τT signifies the timescale at which the beads become relatively more ‘free’ (α(τ) > 0.5) than restricted (α(τ) < 0.5). The progression of τT as a function of Δt is roughly the opposite of αIII, taking the highest values during the period of maximum clot formation between the coagulation and fibrinolysis phases (Fig. 3j). These results suggests that the particles not only exhibit a more restricted dynamical motion (smaller αIII) as the fibrous scaffold is formed around them, but also spend a longer time (larger τT) trapped within their immediate ‘cage’ before being able to explore the surrounding microenvironments.

Our results show that the modulation in αI, α0II, αIII, and τT over the time course of plasma clot formation and fibrinolytic dissolution is more pronounced than the structural variation in fibre density and mesh size. Thus, multiscale analysis of particle dynamics enabled by the laser speckle fluctuation approach may provide a more sensitive markers to characterize the biophysical changes in the particle-fibre network interaction as well as the entrapment and release of particles from the clot scaffolds. As a model of bacterial infection, these nano- to microscale dynamics provide information beyond the typical metrics of structural changes and may offer unique insights into key pathophysiological processes orchestrated by the bacteria and the host immune response.

Discussion

Biological processes in health and disease are regulated by dynamic interactions between cells and the extracellular network of protein scaffolds. Bacterial infection is a prime example in which the interaction between invading pathogens and fibrin(ogen) in the host microenvironment plays an important role in the survival of the pathogens and their ability to establish systemic infection1,2,5. In this work, we present a non-invasive optical approach to investigate the dynamic bacteria-fibrin interactions in a model system of polystyrene microspheres in fibrin scaffolds. Inert microspheres (representing the bacterial cells) have been employed as a model to investigate the behaviour of non-motile bacteria in extracellular matrix scaffold from the biophysical perspective17,18. Based on laser speckle fluctuation, our approach provides access to distinct multiscale dynamics of microspheres in fibrin scaffolds throughout the coagulation cascade in situ. We demonstrate real-time monitoring of particle dynamics during fibrin clot formation and fibrinolytic dissolution in both purified fibrins scaffolds and clinical plasma clots.

Our laser speckle fluctuation approach offers unique capabilities amongst the existing suite of techniques typically utilised in the study of bacteria and bacterial biofilms. Compared to conventional microscopy techniques, our approach uniquely accesses the nano- to microscale dynamical behaviours of the particles that cannot be captured in a snapshot of static morphological images. Our results show that the multiscale dynamics parameters obtained from the timescale-dependent analysis of particle MSD provides a comprehensive set of metrics to characterize biophysical interactions within the particle-fibre network structure at high temporal resolutions that cannot be gleaned from image-based structural metrics alone (Fig. 3). Unlike traditional particle tracking studies, our speckle-based approach does not rely on precise tracking of the position of individual particles, which typically requires high-resolution microscopic imaging or sensitive position detection methods (e.g., quadrant detection). Thus, our approach also circumvents common limitations associated with the use of high-resolution microscopy, including limited applicability in thick (≥60 μm) and turbid samples such as mature bacterial biofilms20. Furthermore, the use of high-speed speckle acquisition to derive the particle MSD enables measurement over a broad range of timescale (on the order of 1 to 10–4 s) at every stage of the coagulation cascade, from unpolymerized fibrinogen solution and plasma to the fully polymerized clots. This broad temporal coverage is critical for capturing the full evolution of particle dynamics that exhibits distinct behaviours across multiple timescales. For instance, traditional particle tracking measurement with a limited frame rate of 10–500 Hz14,18,20,21,22,23,24,25 would typically only observe the plateau behaviour for bound beads in a fully polymerized fibrin clot18, missing potentially valuable insights from the distinct behaviours at shorter timescales revealed by our results.

In this study, we investigated the particle dynamics in fibrin scaffolds and plasma clots containing approximately 5 mg/mL in concentration of fibrinogen, which is within the physiological range in human. Although beyond the scope of the current study, the concentration of fibrinogen is expected to influence both the microstructure of the fibrin scaffolds and the dynamics of the particles within. For instance, whole blood samples with higher fibrinogen content have been found to form clots with greater stiffness and relatively more elastic29,50,51. In fibrin scaffolds, both the clot strength and turbidity have been found to increase with increasing fibrinogen concentration52. The clot elasticity would likely contribute to limiting the extent of particle MSD within the fibrin scaffolds, lowering both the magnitude of the MSD and the α values (particularly α0II, which reflects the viscoelasticity of the fibrin network). Meanwhile, the density of the fibrin network (i.e., mesh size) would limit the extent of particle excursions within the mesh as well as contribute to the steric hindrance that suppresses the values of αI. A sparser fibrin clot could also allow for particles to better escape from the entrapping fibrin scaffolds during fibrinolysis. These concentration effects and their implications on bacteria-fibrin interaction can be investigated using our speckle fluctuation approach and analysis presented in this study.

There are several limitations in the current study that can be addressed for future application of our approach to the study of bacteria and bacterial biofilms. Firstly, the current study is based on a model of polystyrene microspheres of different sizes and surface chemistry to emulate various characteristics of spherical, non-motile bacterial cells. Thus, although these particles may mimic the adhesion binding to fibrin and its impact on the clot microstructures previously observed in bacteria-infected clots, they do not recapitulate the specific surface proteins that decorate the bacterial cell wall or the secretion of bacteria-specific extracellular polymeric substance. Furthermore, fibrin clot formation and fibrinolysis were initiated with human thrombin and tPA rather than the actual bacterial products (e.g., staphylothrombin and staphylokinase secreted by S. aureus), which have been found to produce fibrin clots with abnormal physical properties that may portend systemic dissemination of bacteria as thromboembolic lesions18,19,20. A more clinically relevant model can be implemented in future studies by incorporating actual bacterial cells, which would secrete their own bacterial products, in place of the polystyrene beads. In this case, native light scattering properties of the bacterial population would be directly leveraged for laser speckle fluctuation analysis, and the measurement would be conducted under physiological conditions in an environmental chamber. Alternatively, recombinant bacterial products can be used to initiate coagulation in the particle-fibrin model to investigate the changes in particle dynamics in bacteria-mediated versus normal host-mediated fibrin clots19. In addition, the current study does not consider spatial heterogeneity in the sample, which may be an important factor in the study of bacterial biofilm formation and degradation14,21,22. However, our approach can be adapted to characterize spatially varying particle dynamics via lateral scanning of the illumination beam across the sample. Lastly, the current study only applies to non-motile bacteria with the use of polystyrene microspheres as a model system. However, our approach can be applied to both motile and non-motile bacterial species if actual bacterial cells were to be used as discussed above. Particularly, our analysis of timescale-dependent α parameters can help elucidate the active dynamics (i.e., super-diffusive behaviour) driven by the motility of the bacterial cells20 as they interact with the fibrin scaffolds.

Our laser speckle fluctuation approach provides new means to characterize bacterial behaviour in various models of bacterial infection. Although evidence has shown that bacteria-fibrin interaction is a decisive factor in the progression of infection, how the bacteria behave within the dynamically changing fibrin scaffolds as a result of such interaction is still not well understood. Understanding the evolution of bacteria-fibrin dynamics at the nano- to microscale through the course of coagulation and fibrinolytic events may provide unique insights on the pathogenesis of infection and progression to lethal sepsis. For instance, monitoring how bacterial cells form fibrin-rich aggregates and adhere to implant surfaces may help elucidate the mechanisms for the initiation of periprosthetic joint infection53 and catheter-related infections7. Meanwhile, monitoring how S. aureus hijack host fibrin(ogen) to form protective biofilm scaffolds could help understand the biophysical underpinnings of its immune evasion and antibiotic resistance54. Conversely, dynamical release of entrapped bacteria during fibrinolytic dissolution of the fibrin scaffolds10 may help inform therapeutic strategies to combat biofilm-related bacterial persistence and resistance based on matrix dissolution15,55. Beyond fibrin, our approach is broadly applicable for the investigation of different bacterial strains and their relevant extracellular network scaffolds, including both bacteria-secreted exopolysaccharides and various host proteins.

Methods

Model of bacteria in fibrin scaffolds

S. aureus are spherical non-motile cocci approximately 0.5–1.5 μm in diameter11. The enmeshing of bacteria in fibrin clots is well-emulated by the incorporation of microparticles of similar size into fibrin scaffolds17. Polystyrene microspheres of sizes 0.43, 0.70, and 2.65 μm in diameter were used as models for S. aureus cells. Exact sizes of the microspheres were measured with dynamical light scattering in the backscattered configuration (Malvern, Zetasizer). Due to substantial binding of purified fibrinogen protein to the polystyrene surface, the microspheres were surface functionalized with PEG to block fibrin(ogen) binding. The PEGylation procedure was performed on carboxyl-modified polystyrene microspheres (Bangs Laboratories, PC03001, PC04001, PC05002) via 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDAC) coupling (Bangs Laboratories, PolyLink Protein Coupling Kit), using methoxy PEG reagent with a reactive amine group (Creative PEGWorks, mPEG-Amine, MW 5k)36,41.

Purified fibrin scaffolds (Fig. 2) were prepared with plasminogen-depleted human fibrinogen (Enzyme Research Laboratories, FIB 1), Alexa Fluor 488-conjugated human fibrinogen (Thermo Fisher Scientific, F13191), and human α-thrombin (Enzyme Research Laboratories, HT 1002a) in HBS buffer (20 mM HEPES, 135 mM NaCl, 5 mM CaCl2, pH 7.4) at a final concentration of 5 mg/mL fibrinogen and 2 U/mL thrombin. This fibrinogen concentration is within the physiological content in human plasma, albeit toward the upper end of the normal range. Solutions of fibrinogen and thrombin were prepared 2× the final concentrations in HBS buffer. The unlabelled and fluorescence-labelled fibrinogen were mixed at a molar ratio of 19:1. Polystyrene microspheres were added to the thrombin solution at a final bead volume fraction of 1.8 × 10–3 (0.43 μm), 2.3 × 10–3 (0.70 μm), and 4.5 × 10–3 (2.65 μm), contributing to a reduced scattering coefficient of approximately 0.6 mm–1 in the final mixture. For laser speckle fluctuation measurement, 150 µL each of the fibrinogen and thrombin solutions was added to a 96-well plate and thoroughly mixed by repeated pipetting. For confocal fluorescence microscopy, 75 µL each of the fibrinogen and thrombin solutions was added to a glass bottom Petri dish for microscopy and thoroughly mixed by repeated pipetting.

Plasma clot formation and fibrinolysis

Plasma clots were prepared with discarded clinical whole blood samples obtained from the MGH Core Hematology Laboratory (MGH IRB#2017P000419). Platelet poor plasma was extracted from the whole blood via established laboratory centrifugation protocol. The fibrinogen concentration (5.1 mg/mL) and other clinical coagulation parameters reported by the Core Laboratory are within the normal physiological range. Coagulation was achieved with bovine Q.F.A. thrombin reagent (Werfen, Hemosil, 0020301800) for the Werfen clinical coagulation analyser at a final concentration of 2 U/mL. Fibrinolysis was achieved with tPA (Genentech, Cathflo Activase) at a final concentration of 0.4 μg/mL. Stock tPA was diluted in a solution of 0.2% w/v bovine serum albumin in 0.9% saline. We utilized both the plain polystyrene microspheres (Bangs Laboratories, PS03001) and PEG-coated polystyrene microspheres (Bangs Laboratories, PC03001, PEGylation protocol as described above) with 0.43 μm in diameter to emulate different degrees of fibrin(ogen) binding. The microspheres were added at a final bead volume fraction of 1.1 × 10–3, contributing to a reduced scattering coefficient of approximately 0.4 mm–1 in the final mixture. For laser speckle fluctuation measurement, platelet poor plasma, thrombin solution, tPA solution, and polystyrene microsphere suspensions were mixed in a 96-well plate by repeated pipetting to a total sample volume of 240 µL. For SECM, components were at a total volume of 25 µL, then, a drop was pipetted onto a custom imaging well before the well was covered with a glass coverslip.

Acquisition of laser speckle time series

Laser speckle acquisition system consisted of a fibre-coupled diode laser with wavelength 637 nm (Coherent, OBIS FP 637LX) for illumination and a high-speed CMOS camera (Photron, Mini AX200 type 900k) for speckle acquisition (Supplementary Fig. 4). Speckle time series was acquired via a manufacturer-provided camera control software (Photron FASTCAM Viewer 4). The illumination beam was collimated, linearly polarized, resized to a beam diameter of 1 mm, and focused by an objective lens (convex doublet, focal length 30 mm) to a spot size of 14 μm at the sample. The optical power at the sample was 48 mW. Speckle was collected in a 180° backscattered configuration through an open aperture of 9 mm in diameter. The backscattered light passes through a linear polarizer, oriented perpendicularly to the illumination polarizer to reject specular reflection, before being focused (focal length 400 mm) onto the camera sensor. Samples were positioned below the objective lens with the focal plane just beneath the sample surface. Speckle time series was acquired over a sensor region of 128 × 32 pixels, which allows for a maximum acquisition frame rate of 540 kHz, for a recording duration of at least 1 s. For real-time monitoring of fibrin polymerization (Fig. 2b) and plasma clot formation and dissolution (Fig. 3c, d), speckle time series was repeatedly acquired at the same location at intervals ranging from 22 s (when tracking rapid changes at early phase of polymerization) to 3 min (after samples become more stabilized at late phase of polymerization and fibrinolysis), for a total duration of 50–60 min. For the fully polymerized fibrin clots (Fig. 2g–j), speckle time series was acquired at three different locations in each sample.

Analysis of particle MSD and α from speckle fluctuation

From the raw speckle time series, ensemble-average intensity autocorrelation function, g2(τ), was computed with a contrast-normalized algorithm as described in56:

$${g}_{2}(\tau )=\frac{{g}_{2,\mathrm{raw}}(\tau )-1}{{g}_{2,\mathrm{raw}}(\tau =0)-1}+1;\,{g}_{2,\mathrm{raw}}(\tau )=\frac{\langle I(\tau +{\tau }_{0})I({\tau }_{0})\rangle }{\langle I({\tau }_{0})\rangle \langle I(\tau +{\tau }_{0})\rangle }$$
(1)

where I and τ denote speckle intensity and autocorrelation time, respectively. 〈〉 denotes ensemble average in space (all spatial pixels in a circular region concentric to the illumination centre) and time (all pairs of temporal frames separated by time τ). To ensure optimal speckle contrast in the ensemble, the circular region extends 1/e radius of the diffuse reflectance profile (DRP) obtained by temporally averaging the speckle frames in the time series. Since τ spans several orders of magnitude in timescale, g2(τ) was computed linearly in log-space τ domain with 30 points per decade.

Particle MSD, 〈Δr2(τ)〉, was obtained from the g2(τ) function using our previously derived empirical approximation of the diffusing wave spectroscopy (DWS) formulation26,27:

$${g}_{2}(\tau )-1=\,\exp \left(-2\gamma {\left[{k}^{2}\left\langle {\rm{\Delta }}{r}^{2}(\tau )\right\rangle \right]}^{\zeta }\right)$$
(2)

where k denotes wave number in the medium. Experimental constants γ and ζ account for optical properties of each sample as detailed in prior work26,27. The values of γ and ζ were obtained from a lookup table derived via Monte Carlo ray tracing for a given pair of reduced scattering coefficient and absorption coefficient (negligible for the samples in this study). The optical properties of each sample were estimated by fitting the experimentally measured radial DRP to the photon diffusion theory, as previously described26,27,28,29. The timescale-dependent variation of the MSD can be described with a local power scaling law, α(τ), obtained from the log-log derivative of the MSD with respect to time:

$$\alpha (\tau )=\frac{\partial \,\mathrm{log}\langle \Delta {r}^{2}(\tau )\rangle }{\partial \,\mathrm{log}(\tau )}$$
(3)

In practice, linearly log-space 〈Δr2(τ)〉 (30 points per decade) was first smoothed using weighted linear least squares and a 1st degree polynomial model over a smoothing window size of 7 points. Then, linear regression of log〈Δr2(τ)〉 with respect to log(τ) was computed over a rolling temporal window of width 7 points to obtain α(τ). Finally, α(τ) was smoothed by a moving-average filter with a window size of 11 points.

Multiscale dynamics parameters αI, α0II, αIII, and τT were extracted from a combination of 〈Δr2(τ)〉 and α(τ). In Regime I, αI corresponds to the slope of linear regression of log〈Δr2(τ)〉 with respect to log(τ) over the domain τ < 10–3 s. Confidence interval at 95% of the regression slope is shown by the shaded area in Figs. 2d and 3g. In Regime II, α0II corresponds to the minimum value of α(τ). In Regime III, αIII corresponds to the slope of linear regression of log〈Δr2(τ)〉 with respect to log(τ) over the domain τ > Τ, where Τ is the value of τ at which α(τ) has increased by 50% from the plateau value α0II to the maximum value. Confidence interval at 95% of the regression slope is shown by the shaded area in Figs. 2f and 3i. Lastly, τT corresponds to the value of τ at which α(τ) = 0.5 in Regime III. All data processing and analysis was executed with MATLAB 2022a.

Confocal microscopy

Confocal fluorescence microscopy of the purified fibrin scaffolds (Fig. 2a) was conducted with an Olympus FV1000 confocal laser scanning microscope using a 60× 1.2 NA water immersion objective (Olympus, PLANAPO) and a 488-nm Argon laser for fluorescence excitation. Images were acquired over a field-of-view of 106 μm × 106 μm with 1024 × 1024 pixels. For polymerization monitoring, time-lapsed imaging was conducted with 3 planes (Z increment of 1 μm, starting at ~30 μm from the coverslip) per time point, providing a temporal resolution of ~30 s. Images displayed at top panels of Fig. 2a correspond to a single plane furthest from the coverslip. For fully polymerized clots, Z-stack was acquired over a depth range of 10 μm (Z increment of 0.12 μm, starting at ~30 μm from the coverslip). Images displayed at bottom panels of Fig. 2a correspond to a maximum intensity projection over a depth range of 5 μm closest to the coverslip.

Reflectance confocal microscopy of plasma clots (Fig. 3a, e, f) was conducted with a custom-built SECM system48 using a 100× 1.3 NA oil immersion objective (Zeiss, EC Plan-NeoFluar) and a superluminescent diode (Thorlabs, SLD830S-A20W) for spectrally encoded illumination. The SECM system has a lateral and axial resolution of 0.4 μm and 1 μm, respectively. Time-lapsed SECM images were acquired with 3 planes (Z increment of 1 μm, starting at ~10 μm from the coverslip) per time point over a lateral field-of-view of 2 mm × 0.1 mm, providing a temporal resolution of ~20 s. Images displayed in Fig. 3a correspond to a maximum intensity projection of the three planes. For fully polymerized clots, Z-stack was acquired over a depth range of 30 μm (Z increment of 1 μm, starting at ~10 μm from the coverslip) and a lateral field-of-view of 2 mm × 0.1 mm. Images displayed in Fig. 3e, f correspond to a single plane located ~15 μm from the coverslip.

Statistical analysis

Wilcoxon-type non-parametric test for trend among ordered groups (Cuzick’s test for trend57) was conducted to determine if there was a statistically significant trend in each of the multiscale dynamics parameters, αI, α0II, αIII, across the samples with no bead and increasing bead sizes (Fig. 2g–j). Exact values of the normalized test statistics Z and the associated two-sided P-values are reported; Z > 0 indicates an increasing trend with increasing bead size whereas a Z < 0 indicates a decreasing trend with increasing bead size. Wilcoxon-type non-parametric Kruskal–Wallis test was conducted for multiple comparison of parameters αI, α0II, αIII between any pair of bead sizes (Fig. 2g–j). Significance level of chi-squared statistics on the group-adjusted (Bonferroni correction for multiple comparisons among groups) two-sided pairwise comparison is indicated. Number of measurements were N = 9 (3 measurement locations in each of 3 biological replicates) and P < 0.05 was considered for statistical significance in both tests. Statistical tests were conducted with MATLAB 2022a.