To evaluate the relevant time scale of the formation of a primary clot, we used a setup (Figure 1A) in which the wound was oriented downward and the hemolymph dripped on its own. Caterpillars of 9–10 g and 1–2 days before pre-pupation were used. Each caterpillar was constrained in a tube, and the hemolymph was extracted via a 1–2-mm incision in the third proleg.

This setup allowed us to probe the dripping hemolymph several seconds after making the incision. Filming the features of the drop and liquid bridge formation had a success rate of approximately 5%, because shortly after the incision, the caterpillar moved vigorously, causing difficulties in focusing the camera. Nonetheless, our naked-eye observations and filming confirmed that in the first several seconds after the incision, the detachment of the falling drop from the wound followed the scenario common for simple fluids, such as water (Eggers, 1997; Montanero and Ponce-Torres, 2020); the drop formed a liquid bridge connecting it with the wound (Figure 2A), and then, the bridge necked (Figures 2B, C) and broke, producing satellites (Figures 2D, E). This behavior suggested that in the first seconds, a mostly water-like, low-viscosity hemolymph dripped from the wound.

During the next 5–10 s (i.e., ∼10–15 s after incision), drop detachment showed a different scenario (Figure 3). The necking stage, shown in Figures 2B, C, proceeded similarly (Figures 3A, B), but after that, instead of breaking up the bridge, as shown in Figure 2D, the hemolymph formed filaments of uniform radii (Figure 3C). Before the breakup, the straight filament formed a bulge in the middle (Figure 3D), and the bulge grew to form a second drop (Figures 3E, F). The droplets moved toward the menisci pulled by the connecting filaments. The filaments that formed 10–15 s after the incision broke up in ∼20–40 ms (Figure 3G).

During the next 50–60 s (i.e., ∼75 s after incision), drop detachment followed the scenario shown in Figure 3 (Supplementary Video S1). However, the lifetime of these late filaments increased significantly by approximately one order of magnitude to ∼240 µs.

These results indicate that the properties of the hemolymph change quickly; the hemolymph starts dripping as an almost inviscid water-like fluid and then quickly changes its rheological properties. Typically, bleeding stopped after 60–90 s.

The scenario of wound sealing in the larvae of M. sexta was validated with that of the cockroach P. americana (Figure 4). Although the hemolymph composition of cockroaches could be somewhat different, both the insects have hemocyte-rich hemolymph (Chapman, 2013), and, as shown below, the plasma viscosity for both species is similar. After the detachment of an antennal flagellomere (i.e., an individual unit of the last antennal segment), a drop of hemolymph appeared at the cut end. The drop fell, and the remaining hemolymph moved back into the antennal lumen. No other drops appeared, and after approximately 10 min, the wound had sealed with a rigid crust.

The almost constant radius of the cylindrical filaments indicated that the hemolymph had either thickened by increasing its shear viscosity or had become elastic (McKinley and Sridhar, 2002; Montanero and Ponce-Torres, 2020). In other words, one could think of the filaments as being similar to those either produced by thick honey-like fluids or thin dog saliva-like fluids. To obtain honey-like filaments, the shear viscosity of the hemolymph must increase significantly. We, therefore, measured the shear viscosity by magnetic rotational spectroscopy (MRS) with nickel (Ni) nanorods.

Ferromagnetic nickel nanorods coated with poly(vinylpyrrolidone) were dispersed in a hemolymph drop and were continuously rotated by applying a rotating magnetic field and changing their rotating frequency ω B (Gu and Kornev, 2016; Kornev et al., 2016; Aprelev et al., 2019; Brasovs et al., 2023). When an applied rotating magnetic field B acts on a nanorod of magnetization M and volume V , it exerts a torque, T B = V M B s i n ω B t − θ , forcing the nanorod to turn in the B -plane. The angle ψ = ω B t − θ specifies the angle that the magnetic moment forms with the magnetic field at time moment t . The magnetic moment is aligned with the nanorod axis, forming angle θ with a stationary reference axis (Figure 5).

When a nanorod rotates, it experiences resistance from the surrounding hemolymph. The MRS theory uses the Jeffery (Jeffery, 1922) model of rotation of an ellipsoid in a Newtonian fluid of viscosity η . Therefore, this technique provides the fluid viscosity at the zero shear rate. The torque on the nanorod of length L 0 and diameter d is T η = Γ η d θ / d t , where Γ = π L 0 3 3 ⁡ ln L 0 / d − 2.4 is the nanorod form factor (Doi and Edwards, 1988). For nanorods, the inertial forces at the given frequencies are insignificant; thus, by balancing the mechanical and magnetic torques, T η = T B , we model and interpret the observed behavior and the rotational dynamics of the nanorods in different drops of the hemolymph.

The nanorods were tracked using a camera (Figures 5A–C), and the videos were post-processed to fit the nanorod angle versus time, using this model. Two distinguishable scenarios of nanorod rotation are shown in Figures 5D, E.

At low-driving frequencies, ω B , the nanorod rotated synchronously, lagging behind the field by a constant angle ψ . When the frequency ω B increased above a certain critical value ω c = VMB / ( Γ η ) , the nanorod began to tumble (Figures 6A, B). By measuring ω c , as explained by Kornev et al. (1999); Gu and Kornev (2016); Aprelev et al. (2017); and Brasovs et al. (2023), we measured the shear viscosity of the hemolymph (Table 1). The hemolymph of all probed species showed Newtonian behavior. The shear viscosity of the probed hemolymph did not change over the first few minutes (Figures 6C, D). When probed using a cone-and-plate viscometer at a greater load, the M. sexta hemolymph showed Newtonian behavior as well (Kenny et al., 2018).

This analysis of the shear viscosity suggests that the hemolymph dripping from the wound does not behave as a thick honey-like fluid. Furthermore, the hemocytes had no time to aggregate. Therefore, the appearance of cylindrical filaments when the hemolymph drips from the wound cannot be explained by the significant increase in its shear viscosity.

Thus, we hypothesized that as the first step in primary clot formation, the hemolymph has to be turned into a viscoelastic fluid.

To evaluate the rheological properties of hemolymph filaments, we used the setup shown in Figure 1B. The wound can be probed for only a few minutes after an incision. That is, the time for experimental preparation was long, and when we applied the probe, no dripping occurred. Thus, the wound was partially sealed by the primary clot. Upon placing the probe on the wound and lifting the probe (Figure 1B), we formed the first filament connecting the ball with the wound (Figure 7A). This filament was quite stable; despite the caterpillar’s movements, the meniscus that formed at the filament foot slid across the drop surface and did not break (Figures 7B, C). When the wounded proleg was moved up or down, the filament stretched or contracted, respectively (Figures 7E, H). The first filament took 1–20 s to become thin and break up (Figures 7I, J). We call these liquid bridges “long-lived filaments” (LLFs), indicating that their lifetime exceeded 1 s. These long-lived filaments sometimes developed single or multiple droplets, which remained on the filament until its breakup or moved to the ends. The droplet that moved to the wound was re-absorbed by the wound, and the droplet that moved to the ball was re-absorbed by the film on the ball.

When the probe was applied for the second time, the filament was less stable and broke in less than 1 s (Figure 8). We called these filaments “short-lived filaments” (SLFs). They behaved like those dripping from the wound (Figure 3).

After the second trial, when the probe touched the wound and was then raised, the liquid bridges broke in less than 10 ms, following the breakup stages shown in Figure 2. These observations suggested that the hemolymph behaved as a simple low-viscosity Newtonian liquid, like water, that does not form cylindrical filaments (Eggers, 1997; Montanero and Ponce-Torres, 2020) (see an example in Supplementary Video S2).

Using short-lived filaments as representatives of those dripping from the wound and assuming that they would reflect the properties of the dripping filaments, we characterized their rheological properties. The rheological properties of the hemolymph were evaluated by tracking the radius of the short-lived filaments with time. By analyzing the kinetics of liquid-bridge thinning, we discovered that the bridge radius followed the exponential law R = R 0 ⁡ exp − t 3 θ , (1) with relaxation time θ . These kinetics are typically expected for a viscoelastic fluid following the Oldroyd-B constitutive equation (Bazilevsky et al., 1990; Bazilevskii et al., 2001; McKinley and Sridhar, 2002; Rodd et al., 2005; Kojic et al., 2006; Bazilevsky et al., 2011). The Oldroyd-B fluid is characterized by two constants: its shear viscosity η and relaxation time θ . The measured relaxation time of this viscoelastic hemolymph ranges between 2 and 4 ms (Figure 9). Figure 3 shows that this estimate of the relaxation time is reasonable. Specifically, three filaments were filmed after the drop fell to the substrate, and the filaments thinned in the same manner as that shown in Figure 8 due to capillary action. The analysis of those filaments yielded an average relaxation time of θ ∼ 5   m s . Thus, filament stabilization in the dripping hemolymph can be explained by its viscoelasticity (Larson, 1999; Bazilevskii et al., 2005).

To evaluate the cause of the distinct kinetics of the SLF and LLF thinning, we turned to the structural analysis of their composition. When we probed the wound with the ball in the first 3 min, long-lived filaments were formed 13 times, short-lived filaments were formed 18 times, and 22 trials produced no filaments. For wounds older than 8 min, we observed a long-lived filament only once, while short-lived filaments were observed 11 times, and 52 trials produced no filaments. The long-lived filaments did not necessarily follow the Oldroyd-B model or have a broad dispersion of the relaxation times ranging from milliseconds to seconds. This scattering of the material parameters led us to hypothesize that because of the caterpillar’s movements, the ball might have probed the hemolymph from different depths of ball immersion. We suspected that the ball forming the long-lived filaments might have probed the primary clot together with the layer of hemocytes covering the wound surface. Sometimes, the flowing hemocytes can be observed inside the LLF (Supplementary Video S4).

By collecting long-lived filaments on a glass slide and studying them using phase-contrast microscopy, we discovered filamentous cores made up of polymeric filaments with embedded hemocytes (Figures 9B–D). These filaments were long and bundled, even without hemocytes. When a long-lived filament was broken at one end, some droplets sometimes remained on a dried filament and formed a crusted surface, while the interior of the drop remained full of liquid hemolymph (Figure 9E).

These observations confirm that LLFs contain primary clots, and hence, their thinning kinetics is controlled by the deformation and flow of the highly inhomogeneous composite structure of primary clots. The more mature the primary clot is, the slimier it is, and the slower the filament breaks up.

To further investigate the phenomenon of crust formation, we examined the behavior of hemolymph drops using X-ray micro-CT and SEM imaging. X-ray imaging is attractive because it distinguishes between the structure of the multiphase and multicomponent materials (Andrews et al., 2011; Maire and Withers, 2014). Before submitting the hemolymph drop for X-ray imaging, we rapidly froze the drop on a metal substrate and slowly sublimated the water (as explained in Materials and methods). Then, a freeze-dried drop was imaged using an X-ray microscope (Bruker SkyScan 1176). The images were reconstructed (Figures 10A, B), and we schematically show how the cross-sections parallel and perpendicular to the substrate were taken (Figure 10C). These images show the density distribution of primary aggregates or clots through the drop thickness. The material closer to the air/hemolymph and hemolymph/substrate interfaces had a greater density.

To evaluate the microstructure of this interfacial layer, the frozen drop was broken and removed from the substrate. The remaining material that adhered to the substrate was examined by scanning electron microscopy. Figures 10D–F show residues on the substrate from the droplet sitting on it for longer than 20 min before freezing. Chain-like aggregates are observed attached to the substrate. These aggregates are much larger than the individual hemocytes, suggesting that the hemocytes were already assembled into chain-like primary clots. The possible building blocks of these chain-like aggregates are shown in Figure 10E, where solitary aggregates of approximately 30 μ m in diameter were found. Filamentary residues covered the boundaries of larger primary clots (Figure 10F). These filaments were like those described by Geng and Dunn (1988) and Geng (1990). No visible filaments were found in the droplets incubated for less than 20 min.

All these findings favor the hypothesis that in larval Lepidoptera, the primary clots first nucleate at the interfaces and form a crust-like shell preventing water evaporation. The shell propagates inward to occupy the entire drop.

Hemocytes interact with one another closer to the surface. It has long been hypothesized that some hemocytes extrude thread-like projections (i.e., pseudopodia) in response to foreign surfaces; then, other hemocytes adhere to the formed network (Salt, 1970; Gregoire and Rockstein, 1974). This hypothesis implicitly assumes that the hemocyte density is high enough to allow the connection of two neighboring hemocytes by pseudopodia (Wigglesworth, 1937; Salt, 1970).

We investigated this hypothesis by observing the hemocyte behavior in the droplets released from the wound. Within approximately 3 min from the moment of drop formation, the hemocytes from the larvae of M. sexta began adhering to each other. When a magnetic nanorod was introduced to the drop, the pseudopodia anchored the nanorods to the hemocyte aggregates (Figure 5F); the nanorods never broke the hemocytes. In contrast to the free nanorods, the nanorods attached to hemocyte aggregates did not make full revolutions (Aprelev et al., 2019). Instead, they beat in response to the applied rotating field around an equilibrium direction, revealing hidden interconnections between hemocytes (Supplementary Video S3). During these oscillations, the surrounding cells moved in unison with the nanorod. Thus, the hemocytes with the attached nanorods mechanically communicated with the neighboring hemocytes that were not visibly connected to one another. A hidden biopolymer network provides a far-field reaction to a local perturbation, engaging hemocytes in the movement. This local perturbation of an oscillating nanorod propagates through an area with a radius of ∼5–8 nanorod lengths from the probing nanorod (Figure 11). The analysis of nanorod oscillations in these aggregates suggested that the aggregates with a hidden network that connects them behaved as a viscoelastic material with temporary cross-links (Aprelev et al., 2019), as described by the Green–Tobolsky theory of gels (Green and Tobolsky, 1946).

The high density of hemocytes can only be expected in the neighborhood of the wounded tissue; otherwise, the hemocyte density in a healthy caterpillar is insufficient for hemocyte bridging. We illustrated this statement by showing that after ∼3 min of being close to one another in a large population of hemocytes, only two of them were bridged by a pseudopodium; the other hemocytes remained mobile (Figure 11B).

An illustrative scenario of hemocytes gathering in a primary clot during the first 10 min of hemolymph coagulation is presented in Figure 12 (Aprelev et al., 2019). A pseudopodium extends from the originating mother hemocyte, stretches further, and bends on its way, targeting another hemocyte, to which it later adheres. The pseudopodium functions as a hydraulic spring; on its way to finding the target, it stretches the constituent material. After contact, the pseudopodium tends to spread along the hemocyte surface. The tension generated in the hemocyte during stretching tends to relax, squeezing the constituent material back to the mother hemocyte and dragging the targeted hemocyte with it. As the pseudopodium contracts toward the mother hemocyte, it may encounter another hemocyte, attach to it, and bring it to the mother hemocyte. The process repeats, and the aggregates grow larger and change their shape, leading to the formation of a primary clot (Salt, 1970; Gregoire and Rockstein, 1974; Aprelev et al., 2019).

The mechanism of hemocyte adhesion by pseudopodia can be understood from the analysis of the strength of adhesive bonds between hemocytes in the aggregate. The interfacial tensions and the associated interfacial energies of films separating the adjacent hemocytes offer quantitative metrics for the analysis of aggregation phenomena (Figure 13). These interfacial tensions control the opening angle θ, introduced as the angle of the wedge that the two hemocyte interfaces form at contact (Figure 13D). We distinguish the wedges filled with different media of interest. At least three types of wedges could be formed by the two hemocyte interfaces in the primary aggregate when hemocytes contacted each other (Figures 13A–C). When the wedge is filled with the hemocyte interior, and two interfaces in the wedge separate three different hemocytes, the wedge angle is given as θ = α. When the wedge is filled with plasma, and two interfaces separate hemocytes from the plasma, the wedge angle is given as θ = β. When the wedge is filled with pseudopodia, and two interfaces separate hemocytes from the pseudopodia, the wedge angle is given as θ = γ.

We introduce the interfacial tension σ ⇀ c e l l − c e l l between two identical hemocytes that share a common interface and their interfacial tension σ ⇀ c e l l − m e d i u m separating a hemocyte from the surrounding medium (Figure 13D). These vectors are parallel to interfaces and perpendicular to the contact line that runs perpendicular to the figure. In the introduced Cartesian system of coordinates, the x-axis is parallel to vector σ ⇀ c e l l − c e l l . Setting up a free-body diagram at the vertex, where three interfaces meet (Figure 13D), we have σ ⇀ c e l l − m e d i u m + σ ⇀ c e l l − m e d i u m + σ ⇀ c e l l − c e l l = 0 . (2)

Balancing the x-components of forces, we obtain σ c e l l − m e d i u m ⁡ cos θ 2 + σ c e l l − m e d i u m ⁡ cos θ 2 = σ c e l l − c e l l . (3)

The measured angle α = 120° ± 5° is close to α = 120°, as predicted by Plateau’s law (Plateau, 1873) for the bubble and droplet aggregates with the same interfacial tension σ c e l l − c e l l . Therefore, we conclude that these aggregates are formed by the same hemocytes with the same interfacial energy σ c e l l − c e l l . This interfacial energy σ c e l l − c e l l is assigned for the interfacial energy of hemocytes in the preliminary clot. The relative change in the interfacial energy referenced to σ c e l l − c e l l is introduced as follows: σ m e d i u m * = σ c e l l − m e d i u m σ c e l l − c e l l = 2 ⁡ cos θ 2 − 1 . (4)

Using the measured angles α = 120° ± 5°, β = 70° ± 11°, and γ = 113° ± 6°, the relative interfacial energies of hemocytes in a primary clot ( σ c e l l * ), the two-hemocyte aggregate in the plasma ( σ p l a s m a * ), and the two-hemocyte aggregate in pseudopodia ( σ p s e u d o p o d i a * ) were calculated as σ c e l l * = 1.00 ± 0.08, σ p l a s m a * = 0.61 ± 0.05, and σ p s e u d o p o d i a * = 0.91 ± 0.08, respectively (Figure 13E). These estimates confirm that pseudopodia form a landing bed for incoming hemocytes, and only ∼10% of extra bonds are required to lock the hemocyte in the aggregate.

This analysis emphasizes the importance of interfacial forces in gathering and holding hemocytes together during the development of primary clots.

The larvae of M. sexta were obtained from Carolina Biological Supply (https://www.carolina.com) or were reared in-house on food from Great Lakes Hornworm (https://www.greatlakeshornworm.com/), with some feeding ad libitum on hornworm food obtained from Carolina Biological Supply. The deposited eggs were obtained from adults that emerged in a net enclosure (ca. 27°C and relative humidity ca. 65%). Rearing containers were wide-mouthed 1-L glass jars with strips (ca. 3 × 15 cm) of plastic gutter guard (Frost King Model VX620) as a climbing substrate and food support. Larvae were maintained at controlled room temperature (approximately 25°C) and 24 h of artificial light. To provide gas exchange, the jar lids were replaced with aluminum window screens. Food (ca. 10 mL) was added in the first three instars as needed. In later instars, larvae were removed from jars and placed in clean jars with more food added as needed. The number of larvae per jar was reduced as they grew, with 10 or fewer per jar in the last instars. Larvae entering pre-pupation (i.e., those with a more yellowish thorax, no longer eating, and in a wandering phase) were not used as a source of hemolymph.

The larvae of M. sexta, which were 1–2 days before pre-pupation and weighed more than 8.5 g, were used. The caterpillars were washed free of contaminants with DI water and dried using a paper towel. To minimize the movement of larvae, they were placed in specially designed containers (Figure 1) that gripped the larva along the length of the body while exposing the second and third prolegs. Hemolymph was extracted via a 1–2-mm incision made using a razor blade in the third proleg. The hemolymph exiting the wound was collected on a glass slide or probed directly on the body. All experiments were conducted at 20°C–22°C.

To investigate the LLF structure, a liquid bridge was created between a hemolymph drop and the metal ball (Figure 1C). A glass slide was placed next to the ball perpendicular to the drop surface so that the slide touched the ball’s side. Pulling the ball parallel to the glass slide above the slide edge stretched the liquid bridge and formed a filament. By moving the ball with the filament toward the glass surface, we deposited the filament on the slide. The deposited filament was covered with a cover slip to prevent evaporation during phase-contrast imaging.

To study cell aggregation, drops of blood were placed under an inverted transmitted-light phase-enhanced microscope (Nikon Eclipse Ti; ×40 Oil Objective, S Fluor, NA = 1.30, DIC H/N2; Photometrics CoolSNAP HQ2 camera, 272-ms exposure, time interval = 1 s), and the cell behavior was recorded. Cellular aggregation during the first 10 min of coagulation was observed under high magnification.

A drop of hemolymph was rapidly frozen on a metal block in liquid nitrogen, freeze-dried under vacuum, detached from the metal block, mounted on a low-density Styrofoam block, and imaged using X-ray micro-CT (Bruker SkyScan 1176). The image was digitally reconstructed, and the cross-sections were studied. The methodology can be summarized as follows: an X-ray source in tandem with a detector was rotated around the sample, and projection images of the sample were taken. The images were digitally reconstructed into a three-dimensional structure. The cross-sections of the three-dimensional structure could then be visualized. The resolution of the instrument was 9 μm per pixel, which was inadequate to observe the internal structure of the freeze-dried clots but was sufficient to determine the internal distribution of the material density.

Nickel nanorods of diameter d ∼   200   n m   −   400   n m and length L 0 ∼ 7   μ m − 13   μ m , with a remanent magnetization of M r = 224 kA/m (Aprelev et al., 2017), were dispersed in methanol through sonication. A droplet of the nanorod–methanol suspension was placed on a glass slide and allowed to dry at 23 ±   1 ⁰C and 50 ± 20% humidity. The hemolymph was placed on the dried nanorod–methanol residue and stirred with a glass rod to disperse the nanorods. The hemolymph droplets were directly exposed to air (21–23°C). To eliminate any effects of the hemolymph–substrate and hemolymph–air interfaces on viscosity, only nanorods inside the drop ( 6   μ m below the air–hemolymph interface and ∼   6   μ m above the substrate) were used. The nanorod rotation was filmed, and the videos were processed using LabVIEW and MATLAB algorithms, as explained by Brasovs et al. (2023).

The SLF and LLF breakup was filmed using an IDT Technologies MotionPro X3 camera at 200–900 fps with a resolution of 512–640 pixels and a Grasshopper Point Gray camera at 100–140 fps with a resolution of 1,920–1,200 pixels. The time dependence of the filament radius was obtained with a specially developed LabVIEW Vision Development Module that allowed us to extract and analyze the profile of the liquid bridge for each frame of the videos (Sun et al., 2022; Sun et al., 2023). The module used the IMAQ Extract Contour algorithm to provide the filament profile. Then, the radius of the cylindrical neck of the filament was fit by Eq. 1, and the goodness of fit was evaluated as discussed by Sun et al. (2022) and Sun et al. (2023). The procedure was repeated for each filament, and the average relaxation time for the series was reported.

The LLF hemolymph puddles from 10 caterpillars were tested using the setup shown in Figure 1C. The method required approximately 1.0–1.5 min to collect the hemolymph from the bleeding caterpillar and to prepare the experimental setup. Liquid bridges were formed from the same 5   m m   x   5   m m area of the central spot of the puddles.